Andrew Cooper [Wed, 20 Dec 2017 15:17:26 +0000 (16:17 +0100)]
common/gnttab: Correct error handling for gnttab_setup_table()
Simplify the error labels to just "unlock" and "out". This fixes an erroneous
path where a failure of rcu_lock_domain_by_any_id() still results in
rcu_unlock_domain() being called.
This is only not an XSA by luck. rcu_unlock_domain() is a nop other than
decrementing the preempt count, and nothing reads the preempt count outside of
a debug build.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 5e436e7a45082ea2cadc176c19e1df46c178448f
master date: 2017-08-15 15:08:57 +0100
Jan Beulich [Tue, 12 Dec 2017 14:04:28 +0000 (15:04 +0100)]
x86/shadow: fix ref-counting error handling
The old-Linux handling in shadow_set_l4e() mistakenly ORed together the
results of sh_get_ref() and sh_pin(). As the latter failing is not a
correctness problem, simply ignore its return value.
In sh_set_toplevel_shadow() a failing sh_get_ref() must not be
accompanied by installing the entry, despite the domain being crashed.
This is XSA-250.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Tim Deegan <tim@xen.org>
master commit: 10be8001de7d87be1f0ccdda75cc70e922e56d03
master date: 2017-12-12 14:29:45 +0100
Jan Beulich [Tue, 12 Dec 2017 14:04:00 +0000 (15:04 +0100)]
x86/shadow: fix refcount overflow check
Commit c385d27079 ("x86 shadow: for multi-page shadows, explicitly track
the first page") reduced the refcount width to 25, without adjusting the
overflow check. Eliminate the disconnect by using a manifest constant.
Interestingly, up to commit 047782fa01 ("Out-of-sync L1 shadows: OOS
snapshot") the refcount was 27 bits wide, yet the check was already
using 26.
This is XSA-249.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com> Reviewed-by: Tim Deegan <tim@xen.org>
master commit: 54e2292e8df7a1a7b041192be9d6d797b6d00869
master date: 2017-12-12 14:29:13 +0100
Jan Beulich [Tue, 12 Dec 2017 14:03:34 +0000 (15:03 +0100)]
x86/mm: don't wrongly set page ownership
PV domains can obtain mappings of any pages owned by the correct domain,
including ones that aren't actually assigned as "normal" RAM, but used
by Xen internally. At the moment such "internal" pages marked as owned
by a guest include pages used to track logdirty bits, as well as p2m
pages and the "unpaged pagetable" for HVM guests. Since the PV memory
management and shadow code conflict in their use of struct page_info
fields, and since shadow code is being used for log-dirty handling for
PV domains, pages coming from the shadow pool must, for PV domains, not
have the domain set as their owner.
While the change could be done conditionally for just the PV case in
shadow code, do it unconditionally (and for consistency also for HAP),
just to be on the safe side.
There's one special case though for shadow code: The page table used for
running a HVM guest in unpaged mode is subject to get_page() (in
set_shadow_status()) and hence must have its owner set.
This is XSA-248.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Tim Deegan <tim@xen.org> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
master commit: ff2a793e15bb0b6254bc849ef8e83e1c284c3583
master date: 2017-12-12 14:28:36 +0100
Jan Beulich [Tue, 12 Dec 2017 14:03:00 +0000 (15:03 +0100)]
x86: don't wrongly trigger linear page table assertion (2)
_put_final_page_type(), when free_page_type() has exited early to allow
for preemption, should not update the time stamp, as the page continues
to retain the typ which is in the process of being unvalidated. I can't
see why the time stamp update was put on that path in the first place
(albeit it may well have been me who had put it there years ago).
This is part of XSA-240.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Tested-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: George Dunlap <george.dunlap.com>
master commit: e40b0219a8c77741ae48989efb520f4a762a5be3
master date: 2017-12-12 14:27:34 +0100
George Dunlap [Tue, 28 Nov 2017 12:45:13 +0000 (13:45 +0100)]
p2m: Check return value of p2m_set_entry() when decreasing reservation
If the entire range specified to p2m_pod_decrease_reservation() is marked
populate-on-demand, then it will make a single p2m_set_entry() call,
reducing its PoD entry count.
Unfortunately, in the right circumstances, this p2m_set_entry() call
may fail. It that case, repeated calls to decrease_reservation() may
cause p2m->pod.entry_count to fall below zero, potentially tripping
over BUG_ON()s to the contrary.
Instead, check to see if the entry succeeded, and return false if not.
The caller will then call guest_remove_page() on the gfns, which will
return -EINVAL upon finding no valid memory there to return.
Unfortunately if the order > 0, the entry may have partially changed.
A domain_crash() is probably the safest thing in that case.
Other p2m_set_entry() calls in the same function should be fine,
because they are writing the entry at its current order. Nonetheless,
check the return value and crash if our assumption turns otu to be
wrong.
This is part of XSA-247.
Signed-off-by: George Dunlap <george.dunlap@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: a3d64de8e86f5812917d2d0af28298f80debdf9a
master date: 2017-11-28 13:13:26 +0100
George Dunlap [Tue, 28 Nov 2017 12:44:42 +0000 (13:44 +0100)]
p2m: Always check to see if removing a p2m entry actually worked
The PoD zero-check functions speculatively remove memory from the p2m,
then check to see if it's completely zeroed, before putting it in the
cache.
Unfortunately, the p2m_set_entry() calls may fail if the underlying
pagetable structure needs to change and the domain has exhausted its
p2m memory pool: for instance, if we're removing a 2MiB region out of
a 1GiB entry (in the p2m_pod_zero_check_superpage() case), or a 4k
region out of a 2MiB or larger entry (in the p2m_pod_zero_check()
case); and the return value is not checked.
The underlying mfn will then be added into the PoD cache, and at some
point mapped into another location in the p2m. If the guest
afterwards ballons out this memory, it will be freed to the hypervisor
and potentially reused by another domain, in spite of the fact that
the original domain still has writable mappings to it.
There are several places where p2m_set_entry() shouldn't be able to
fail, as it is guaranteed to write an entry of the same order that
succeeded before. Add a backstop of crashing the domain just in case,
and an ASSERT_UNREACHABLE() to flag up the broken assumption on debug
builds.
While we're here, use PAGE_ORDER_2M rather than a magic constant.
This is part of XSA-247.
Signed-off-by: George Dunlap <george.dunlap@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 92790672dedf2eab042e04ecc277c19d40fd348a
master date: 2017-11-28 13:13:03 +0100
Julien Grall [Tue, 28 Nov 2017 12:44:09 +0000 (13:44 +0100)]
x86/pod: prevent infinite loop when shattering large pages
When populating pages, the PoD may need to split large ones using
p2m_set_entry and request the caller to retry (see ept_get_entry for
instance).
p2m_set_entry may fail to shatter if it is not possible to allocate
memory for the new page table. However, the error is not propagated
resulting to the callers to retry infinitely the PoD.
Prevent the infinite loop by return false when it is not possible to
shatter the large mapping.
This is XSA-246.
Signed-off-by: Julien Grall <julien.grall@linaro.org> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
master commit: a1c6c6768971ea387d7eba0803908ef0928b43ac
master date: 2017-11-28 13:11:55 +0100
Andrew Cooper [Thu, 16 Nov 2017 11:03:26 +0000 (12:03 +0100)]
x86/shadow: correct SH_LINEAR mapping detection in sh_guess_wrmap()
The fix for XSA-243 / CVE-2017-15592 (c/s bf2b4eadcf379) introduced a change
in behaviour for sh_guest_wrmap(), where it had to cope with no shadow linear
mapping being present.
As the name suggests, guest_vtable is a mapping of the guests pagetable, not
Xen's pagetable, meaning that it isn't the pagetable we need to check for the
shadow linear slot in.
The practical upshot is that a shadow HVM vcpu which switches into 4-level
paging mode, with an L4 pagetable that contains a mapping which aliases Xen's
SH_LINEAR_PT_VIRT_START will fool the safety check for whether a SHADOW_LINEAR
mapping is present. As the check passes (when it should have failed), Xen
subsequently falls over the missing mapping with a pagefault such as:
Jan Beulich [Thu, 16 Nov 2017 11:02:55 +0000 (12:02 +0100)]
x86: don't wrongly trigger linear page table assertion
_put_page_type() may do multiple iterations until its cmpxchg()
succeeds. It invokes set_tlbflush_timestamp() on the first
iteration, however. Code inside the function takes care of this, but
- the assertion in _put_final_page_type() would trigger on the second
iteration if time stamps in a debug build are permitted to be
sufficiently much wider than the default 6 bits (see WRAP_MASK in
flushtlb.c),
- it returning -EINTR (for a continuation to be scheduled) would leave
the page inconsistent state (until the re-invocation completes).
Make the set_tlbflush_timestamp() invocation conditional, bypassing it
(for now) only in the case we really can't tolerate the stamp to be
stored.
This is part of XSA-240.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
master commit: 2c458dfcb59f3d9d8a35fc5ffbf780b6ed7a26a6
master date: 2017-11-16 10:37:29 +0100
Yu Zhang [Thu, 16 Nov 2017 11:02:24 +0000 (12:02 +0100)]
x86/mm: fix race condition in modify_xen_mappings()
In modify_xen_mappings(), a L1/L2 page table shall be freed,
if all entries of this page table are empty. Corresponding
L2/L3 PTE will need be cleared in such scenario.
However, concurrent paging structure modifications on different
CPUs may cause the L2/L3 PTEs to be already be cleared or set
to reference a superpage.
Therefore the logic to enumerate the L1/L2 page table and to
reset the corresponding L2/L3 PTE need to be protected with
spinlock. And the _PAGE_PRESENT and _PAGE_PSE flags need be
checked after the lock is obtained.
Suggested-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: b9ee1fd7b98064cf27d0f8f1adf1f5359b72c97f
master date: 2017-11-14 17:11:26 +0100
Min He [Thu, 16 Nov 2017 11:01:57 +0000 (12:01 +0100)]
x86/mm: fix race conditions in map_pages_to_xen()
In map_pages_to_xen(), a L2 page table entry may be reset to point to
a superpage, and its corresponding L1 page table need be freed in such
scenario, when these L1 page table entries are mapping to consecutive
page frames and having the same mapping flags.
However, variable `pl1e` is not protected by the lock before L1 page table
is enumerated. A race condition may happen if this code path is invoked
simultaneously on different CPUs.
For example, `pl1e` value on CPU0 may hold an obsolete value, pointing
to a page which has just been freed on CPU1. Besides, before this page
is reused, it will still be holding the old PTEs, referencing consecutive
page frames. Consequently the `free_xen_pagetable(l2e_to_l1e(ol2e))` will
be triggered on CPU0, resulting the unexpected free of a normal page.
This patch fixes the above problem by protecting the `pl1e` with the lock.
Also, there're other potential race conditions. For instance, the L2/L3
entry may be modified concurrently on different CPUs, by routines such as
map_pages_to_xen(), modify_xen_mappings() etc. To fix this, this patch will
check the _PAGE_PRESENT and _PAGE_PSE flags, after the spinlock is obtained,
for the corresponding L2/L3 entry.
Signed-off-by: Min He <min.he@intel.com> Signed-off-by: Yi Zhang <yi.z.zhang@intel.com> Signed-off-by: Yu Zhang <yu.c.zhang@linux.intel.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: a5114662297ad03efc36b52ad365ffa05fb357b7
master date: 2017-11-14 17:10:56 +0100
Eric Chanudet [Thu, 16 Nov 2017 11:01:28 +0000 (12:01 +0100)]
x86/hvm: do not register hpet mmio during s3 cycle
Do it once at domain creation (hpet_init).
Sleep -> Resume cycles will end up crashing an HVM guest with hpet as
the sequence during resume takes the path:
-> hvm_s3_suspend
-> hpet_reset
-> hpet_deinit
-> hpet_init
-> register_mmio_handler
-> hvm_next_io_handler
register_mmio_handler will use a new io handler each time, until
eventually it reaches NR_IO_HANDLERS, then hvm_next_io_handler calls
domain_crash.
Signed-off-by: Eric Chanudet <chanudete@ainfosec.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 015d6738ddff4074668c1d4887bbffd507ed1a7f
master date: 2017-11-14 17:09:50 +0100
George Dunlap [Thu, 16 Nov 2017 11:00:28 +0000 (12:00 +0100)]
x86/mm: Make PV linear pagetables optional
Allowing pagetables to point to other pagetables of the same level
(often called 'linear pagetables') has been included in Xen since its
inception; but recently it has been the source of a number of subtle
reference-counting bugs.
It is not used by Linux or MiniOS; but it is used by NetBSD and Novell
Netware. There are significant numbers of people who are never going
to use the feature, along with significant numbers who need the
feature.
Add a Kconfig option for the feature (default to 'y'). Also add a
command-line option to control whether PV linear pagetables are
allowed (default to 'true').
NB that we leave linear_pt_count in the page struct. It's in a union,
so its presence doesn't increase the size of the data struct.
Changing the layout of the other elements based on configuration
options is asking for trouble however; so we'll just leave it there
and ASSERT that it's zero.
Reported-by: Jann Horn <jannh@google.com> Signed-off-by: George Dunlap <george.dunlap@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 3285e75dea89afb0ef5b3ee39bd15194bd7cc110
master date: 2017-10-27 14:36:45 +0100
Jan Beulich [Thu, 16 Nov 2017 10:59:21 +0000 (11:59 +0100)]
x86: don't latch wrong (stale) GS base addresses
load_segments() writes selector registers before doing any of the base
address updates. Any of these selector loads can cause a page fault in
case it references the LDT, and the LDT page accessed was only recently
installed. Therefore the call tree map_ldt_shadow_page() ->
guest_get_eff_kern_l1e() -> toggle_guest_mode() would in such a case
wrongly latch the outgoing vCPU's GS.base into the incoming vCPU's
recorded state.
Split page table toggling from GS handling - neither
guest_get_eff_kern_l1e() nor guest_io_okay() need more than the page
tables being the kernel ones for the memory access they want to do.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: a711f6f24a7157ae70d1cc32e61b98f23dc0c584
master date: 2017-10-27 13:49:10 +0100
Jan Beulich [Thu, 16 Nov 2017 10:58:47 +0000 (11:58 +0100)]
x86: also show FS/GS base addresses when dumping registers
Their state may be important to figure the reason for a crash. To not
further grow duplicate code, break out a helper function.
I realize that (ab)using the control register array here may not be
considered the nicest solution, but it seems easier (and less overall
overhead) to do so compared to the alternative of introducing another
helper structure.
Signed-off-by: Jan Beulich <jbeulich@suse.com>
Reviewed-by Andrew Cooper <andrew.cooper3@citrix.com>
master commit: be7f60b5a39741eab0a8fea0324f7be0cb724cfb
master date: 2017-10-24 18:13:13 +0200
Jan Beulich [Thu, 16 Nov 2017 10:58:10 +0000 (11:58 +0100)]
x86: fix GS-base-dirty determination
load_segments() writes the two MSRs in their "canonical" positions
(GS_BASE for the user base, SHADOW_GS_BASE for the kernel one) and uses
SWAPGS to switch them around if the incoming vCPU is in kernel mode. In
order to not leave a stale kernel address in GS_BASE when the incoming
guest is in user mode, the check on the outgoing vCPU needs to be
dependent upon the mode it is currently in, rather than blindly looking
at the user base.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 91f85280b9b80852352fcad73d94ed29fafb88da
master date: 2017-10-24 18:12:31 +0200
Jan Beulich [Tue, 24 Oct 2017 14:50:14 +0000 (16:50 +0200)]
x86emul: handle address wrapping
This just the emulator part of commit 7869e2bafe
("x86emul/fuzz: add rudimentary limit checking"):
Several adjustments to the emulator's address calculations are needed:
While the DstBitBase one is really mandatory, the specification allows
for either original or new behavior for two-part accesses. Observed
behavior on real hardware, however, is for such accesses to silently
wrap at the 2^^32 boundary in other than 64-bit mode, just like they do
at the 2^^64 boundary in 64-bit mode, which our code is now being
brought in line with. While adding truncate_ea() invocations there,
also convert open coded instances of it.
Reported-by: George Dunlap <george.dunlap@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Tue, 24 Oct 2017 14:48:50 +0000 (16:48 +0200)]
x86: avoid #GP for PV guest MSR accesses
Halfway recent Linux kernels probe MISC_FEATURES_ENABLES on all CPUs,
leading to ugly recovered #GP fault messages with debug builds on older
systems. We can do better, so introduce synthetic feature flags for
both this and PLATFORM_INFO to avoid the rdmsr_safe() altogether.
Note that the r/o nature of PLATFORM_INFO is now also being enforced.
The rdmsr_safe() uses for MISC_ENABLE are left in place as benign - it
exists for all 64-bit capable Intel CPUs (see e.g. early_init_intel()).
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Andrew Cooper [Tue, 24 Oct 2017 14:43:43 +0000 (16:43 +0200)]
x86/vvmx: Fix WRMSR interception of VMX MSRs
FEATURE_CONTROL is already read with LOCK bit set (so is unmodifiable), and
all VMX MSRs are read-only. Also, fix the MSR_IA32_VMX_TRUE_ENTRY_CTLS bound
to be MSR_IA32_VMX_VMFUNC, rather than having the intervening MSRs falling
into the default case.
Raise #GP faults if the guest tries to modify any of them.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Acked-by: Kevin Tian <kevin.tian@intel.com>
master commit: 46c3acb308bf0cd044b114e637aacaf18b957618
master date: 2017-06-30 11:27:50 +0100
While I can't seem to find any users of this hypercall (being a likely
explanation of why the problem wasn't noticed so far), just like for
do_mmu_update() paged-out and shared page handling is needed here. Move
all this logic into mod_l1_entry(), which then also results in no
longer
- doing any of this handling for non-present PTEs,
- acquiring two temporary page references when one is already more than
enough.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 46aaf41ee099a048d7a554c03ae01bcdaa07f776
master date: 2017-10-13 12:43:41 +0200
Andrew Cooper [Tue, 24 Oct 2017 14:42:13 +0000 (16:42 +0200)]
xen/domctl: Fix Xen heap leak via XEN_DOMCTL_getvcpucontext
The backing structure for XEN_DOMCTL_getvcpucontext is only zeroed in the x86
HVM case. At the very least, this means that ARM returns junk through its
flags field (as it is only ever conditionally or'd into), and x86 PV leaks
data through gdt_frames[14...15]. (An exhaustive search for other leaks
hasn't been performed).
Unconditionally zero the memory upon allocation, and forgo the double clear
for x86 HVM. These hypercalls are not on hotpaths.
Note that this does not qualify for an XSA. Per XSA-77,
XEN_DOMCTL_getvcpucontext is unsafe for disaggregation, meaning that only the
control domain can use this hypercall.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 3b2eeb7412e529f38d1e8b872ba0bc6ab09a7008
master date: 2017-10-09 12:43:21 +0100
Segment bases (and limits) aren't being cleared by the loading of a nul
selector into a segment register on AMD CPUs. Therefore, if an
outgoing vCPU has a non-zero base in FS or GS and the subsequent
incoming vCPU has a non-zero but nul selector in the respective
register(s), the selector value(s) would be loaded without clearing the
segment base(s) in the hidden register portion.
Since the ABI states "zero" in its description of the fs and gs fields,
it is worth noting that the chosen approach to fix this alters the
written down ABI. I consider this preferrable over enforcing the
previously written down behavior, as nul selectors are far more likely
to be what was meant from the beginning.
The adjustments also eliminate an inconsistency between FS and GS
handling: Old code had an extra pointless (gs_base_user was always zero
when DIRTY_GS was set) conditional for GS. The old bitkeeper changeset
has no explanation for this asymmetry.
Additionally for DS and ES a flat selector is being loaded prior to the
loading of a nul one on AMD CPUs, just as a precautionary measure
(we're not currently aware of ways for a guest to deduce the base of a
segment register which has a nul selector loaded).
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 4e383df8650d72e47e2ca4ebfc4f6986f791d2f2
master date: 2017-10-04 14:17:08 +0200
Andrew Cooper [Tue, 24 Oct 2017 14:40:23 +0000 (16:40 +0200)]
x86/svm: Fix a livelock when trying to run shadowed unpaged guests
On AMD processors which support SMEP (Some Fam16h processors) and SMAP (Zen,
Fam17h), a guest which is running with shadow paging and clears CR0.PG while
keeping CR4.{SMEP,SMAP} set will livelock, as hardware raises #PF which the
shadow pagetable concludes shouldn't happen.
This occurs because hardware is running with host paging settings, which
causes the guests choice of SMEP/SMAP to actually take effect, even though
they shouldn't from the guests point of view.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
master commit: 3164f2f9db1e63ea64c3f9520d40cb09920d2b35
master date: 2017-10-02 13:57:34 +0100
Jan Beulich [Tue, 24 Oct 2017 14:39:33 +0000 (16:39 +0200)]
gnttab: fix pin count / page reference race
Dropping page references before decrementing pin counts is a bad idea
if assumptions are being made that a non-zero pin count implies a valid
page. Fix the order of operations in gnttab_copy_release_buf(), but at
the same time also remove the assertion that was found to trigger:
map_grant_ref() also has the potential of causing a race here, and
changing the order of operations there would likely be quite a bit more
involved.
This is CVE-2017-15597 / XSA-236.
Reported-by: Pawel Wieczorkiewicz <wipawel@amazon.de> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: e008f7619dcd6d549727c9635b3f9f3c7ee483ed
master date: 2017-10-24 16:01:33 +0200
The variable domctl.u.address_size.size may remain uninitialized if
guest_type is not one of xen-3.0-aarch64 or xen-3.0-armv7l. And the
code precisely checks if this variable is still 0 to decide if the
guest type is supported or not.
This fixes the following build failure with gcc 7.x:
xc_dom_arm.c:229:31: error: 'domctl.u.address_size.size' may be used uninitialized in this function [-Werror=maybe-uninitialized]
if ( domctl.u.address_size.size == 0 )
Patch originally taken from
https://www.mail-archive.com/xen-devel@lists.xen.org/msg109313.html.
Signed-off-by: Bernd Kuhls <bernd.kuhls@t-online.de> Signed-off-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
(cherry picked from commit 88bfbf90e35f1213f9967a97dee0b2039f9998a4)
Andrew Cooper [Thu, 12 Oct 2017 13:30:21 +0000 (15:30 +0200)]
x86/cpu: Fix IST handling during PCPU bringup
Clear IST references in newly allocated IDTs. Nothing good will come of
having them set before the TSS is suitably constructed (although the chances
of the CPU surviving such an IST interrupt/exception is extremely slim).
Uniformly set the IST references after the TSS is in place. This fixes an
issue on AMD hardware, where onlining a PCPU while PCPU0 is in HVM context
will cause IST_NONE to be copied into the new IDT, making that PCPU vulnerable
to privilege escalation from PV guests until it subsequently schedules an HVM
guest.
This is XSA-244.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: cc08c73c8c1f5ba5ed0f8274548db6725e1c3157
master date: 2017-10-12 14:50:31 +0200
Andrew Cooper [Thu, 12 Oct 2017 13:29:52 +0000 (15:29 +0200)]
x86/shadow: Don't create self-linear shadow mappings for 4-level translated guests
When initially creating a monitor table for 4-level translated guests, don't
install a shadow-linear mapping. This mapping is actually self-linear, and
trips up the writeable heuristic logic into following Xen's mappings, not the
guests' shadows it was expecting to follow.
A consequence of this is that sh_guess_wrmap() needs to cope with there being
no shadow-linear mapping present, which in practice occurs once each time a
vcpu switches to 4-level paging from a different paging mode.
An appropriate shadow-linear slot will be inserted into the monitor table
either while constructing lower level monitor tables, or by sh_update_cr3().
While fixing this, clarify the safety of the other mappings. Despite
appearing unsafe, it is correct to create a guest-linear mapping for
translated domains; this is self-linear and doesn't point into the translated
domain. Drop a dead clause for translate != external guests.
This is XSA-243.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Tim Deegan <tim@xen.org>
master commit: bf2b4eadcf379d0361b38de9725ea5f7a18a5205
master date: 2017-10-12 14:50:07 +0200
Jan Beulich [Thu, 12 Oct 2017 13:29:21 +0000 (15:29 +0200)]
x86: don't allow page_unlock() to drop the last type reference
Only _put_page_type() does the necessary cleanup, and hence not all
domain pages can be released during guest cleanup (leaving around
zombie domains) if we get this wrong.
Jan Beulich [Thu, 12 Oct 2017 13:28:44 +0000 (15:28 +0200)]
x86: don't store possibly stale TLB flush time stamp
While the timing window is extremely narrow, it is theoretically
possible for an update to the TLB flush clock and a subsequent flush
IPI to happen between the read and write parts of the update of the
per-page stamp. Exclude this possibility by disabling interrupts
across the update, preventing the IPI to be serviced in the middle.
This is XSA-241.
Reported-by: Jann Horn <jannh@google.com> Suggested-by: George Dunlap <george.dunlap@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
master commit: 23a183607a427572185fc51c76cc5ab11c00c4cc
master date: 2017-10-12 14:48:25 +0200
Jan Beulich [Thu, 12 Oct 2017 13:27:37 +0000 (15:27 +0200)]
x86: limit linear page table use to a single level
That's the only way that they're meant to be used. Without such a
restriction arbitrarily long chains of same-level page tables can be
built, tearing down of which may then cause arbitrarily deep recursion,
causing a stack overflow. To facilitate this restriction, a counter is
being introduced to track both the number of same-level entries in a
page table as well as the number of uses of a page table in another
same-level one (counting into positive and negative direction
respectively, utilizing the fact that both counts can't be non-zero at
the same time).
Note that the added accounting introduces a restriction on the number
of times a page can be used in other same-level page tables - more than
32k of such uses are no longer possible.
Note also that some put_page_and_type[_preemptible]() calls are
replaced with open-coded equivalents. This seemed preferrable to
adding "parent_table" to the matrix of functions.
Note further that cross-domain same-level page table references are no
longer permitted (they probably never should have been).
This is XSA-240.
Reported-by: Jann Horn <jannh@google.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: George Dunlap <george.dunlap@citrix.com>
master commit: 6987fc7558bdbab8119eabf026e3cdad1053f0e5
master date: 2017-10-12 14:44:34 +0200
Jan Beulich [Thu, 12 Oct 2017 13:27:07 +0000 (15:27 +0200)]
x86/HVM: prefill partially used variable on emulation paths
Certain handlers ignore the access size (vioapic_write() being the
example this was found with), perhaps leading to subsequent reads
seeing data that wasn't actually written by the guest. For
consistency and extra safety also do this on the read path of
hvm_process_io_intercept(), even if this doesn't directly affect what
guests get to see, as we've supposedly already dealt with read handlers
leaving data completely unitialized.
Misbehaving device model can pass incorrect XEN_DMOP_map/
unmap_io_range_to_ioreq_server arguments, namely end < start when
specifying address range. When this happens we hit ASSERT(s <= e) in
rangeset_contains_range()/rangeset_overlaps_range() with debug builds.
Production builds will not trap right away but may misbehave later
while handling such bogus ranges.
Jan Beulich [Thu, 12 Oct 2017 13:25:57 +0000 (15:25 +0200)]
x86/FLASK: fix unmap-domain-IRQ XSM hook
The caller and the FLASK implementation of xsm_unmap_domain_irq()
disagreed about what the "data" argument points to in the MSI case:
Change both sides to pass/take a PCI device.
This is part of XSA-237.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 6f17f5c43a3bd28d27ed8133b2bf513e2eab7d59
master date: 2017-10-12 14:37:56 +0200
Jan Beulich [Thu, 12 Oct 2017 13:24:35 +0000 (15:24 +0200)]
x86/MSI: disallow redundant enabling
At the moment, Xen attempts to allow redundant enabling of MSI by
having pci_enable_msi() return 0, and point to the existing MSI
descriptor, when the msi already exists.
Unfortunately, if subsequent errors are encountered, the cleanup
paths assume pci_enable_msi() had done full initialization, and
hence undo everything that was assumed to be done by that
function without also undoing other setup that would normally
occur only after that function was called (in map_domain_pirq()
itself).
Rather than try to make the redundant enabling case work properly, just
forbid it entirely by having pci_enable_msi() return -EEXIST when MSI
is already set up.
This is part of XSA-237.
Reported-by: HW42 <hw42@ipsumj.de> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
master commit: a46126fec20e0cf4f5442352ef45efaea8c89646
master date: 2017-10-12 14:36:58 +0200
Jan Beulich [Thu, 12 Oct 2017 13:24:03 +0000 (15:24 +0200)]
x86: enforce proper privilege when (un)mapping pIRQ-s
(Un)mapping of IRQs, just like other RESOURCE__ADD* / RESOURCE__REMOVE*
actions (in FLASK terms) should be XSM_DM_PRIV rather than XSM_TARGET.
This in turn requires bypassing the XSM check in physdev_unmap_pirq()
for the HVM emuirq case just like is being done in physdev_map_pirq().
The primary goal security wise, however, is to no longer allow HVM
guests, by specifying their own domain ID instead of DOMID_SELF, to
enter code paths intended for PV guest and the control domains of HVM
guests only.
This is part of XSA-237.
Reported-by: HW42 <hw42@ipsumj.de> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
master commit: db72faf69c94513e180568006a9d899ed422ff90
master date: 2017-10-12 14:36:30 +0200
Jan Beulich [Thu, 12 Oct 2017 13:23:20 +0000 (15:23 +0200)]
x86: don't allow MSI pIRQ mapping on unowned device
MSI setup should be permitted only for existing devices owned by the
respective guest (the operation may still be carried out by the domain
controlling that guest).
This is part of XSA-237.
Reported-by: HW42 <hw42@ipsumj.de> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 3308374b1be7d43e23bd2e9eaf23ec06d7959882
master date: 2017-10-12 14:35:14 +0200
Julien Grall [Fri, 6 Oct 2017 13:16:50 +0000 (15:16 +0200)]
xen/arm: Correctly report the memory region in the dummy NUMA helpers
NUMA is currently not supported on Arm. Because common code is
NUMA-aware, dummy helpers are instead provided to expose a single node.
Those helpers are for instance used to know the region to scrub.
However the memory region is not reported correctly. Indeed, the
frametable may not be at the beginning of the memory and there might be
multiple memory banks. This will lead to not scrub some part of the
memory.
The memory information can be found using:
* first_valid_mfn as the start of the memory
* max_page - first_valid_mfn as the spanned pages
Note that first_valid_mfn is now been exported. The prototype has been
added in asm-arm/numa.h and not in a common header because I would
expect the variable to become static once NUMA is fully supported on
Arm.
Julien Grall [Fri, 6 Oct 2017 13:16:12 +0000 (15:16 +0200)]
xen/page_alloc: Cover memory unreserved after boot in first_valid_mfn
On Arm, some regions (e.g Initramfs, Dom0 Kernel...) are marked as
reserved until the hardware domain is built and they are copied into its
memory. Therefore, they will not be added in the boot allocator via
init_boot_pages.
Instead, init_xenheap_pages will be called once the region are not used
anymore.
Update first_valid_mfn in both init_heap_pages and init_boot_pages
(already exist) to cover all the cases.
Chao Gao [Fri, 6 Oct 2017 13:14:47 +0000 (15:14 +0200)]
VT-d: use correct BDF for VF to search VT-d unit
When SR-IOV is enabled, 'Virtual Functions' of a 'Physical Function'
are under the scope of the same VT-d unit as the 'Physical Function'.
A 'Physical Function' can be a 'Traditional Function' or an ARI
'Extended Function'. And furthermore, 'Extended Functions' on an
endpoint are under the scope of the same VT-d unit as the 'Traditional
Functions' on the endpoint. To search VT-d unit for a VF, if its PF
isn't an extended function, the BDF of PF should be used. Otherwise
the BDF of a traditional function in the same device with the PF
should be used.
Current code uses PCI_SLOT() to recognize an ARI 'Extended Funcion'.
But it is conceptually wrong w/o checking whether PF is an extended
function and would lead to match VFs of a RC integrated PF to a wrong
VT-d unit.
This patch overrides VF 'is_extfn' field and uses this field to
indicate whether the PF of this VF is an extended function. The field
helps to use correct BDF to search VT-d unit.
Reported-by: Crawford, Eric R <Eric.R.Crawford@intel.com> Signed-off-by: Chao Gao <chao.gao@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Acked-by: Jan Beulich <jbeulich@suse.com> Tested-by: Crawford, Eric R <Eric.R.Crawford@intel.com>
master commit: c286af54c7177c14180121b422d8df7281e547cb
master date: 2017-09-01 11:02:23 +0200
Xiong Zhang [Fri, 6 Oct 2017 13:14:20 +0000 (15:14 +0200)]
hvmloader: use base instead of pci_mem_start for find_next_rmrr()
find_next_rmrr(base) is used to find the lowest RMRR ending above base
but below 4G. Current method couldn't cover the following situation:
a. two rmrr exist, small gap between them
b. pci_mem_start and mem_resource.base is below the first rmrr.base
c. find_next_rmrr(pci_mem_start) will find the first rmrr
d. After aligning mem_resource.base to bar size,
first_rmrr.end < new_base < second_rmrr.base and
new_base + bar_sz > second_rmrr.base.
So the new bar will overlap with the second rmrr and doesn't overlap
with the first rmrr.
But the next_rmrr point to the first rmrr, then check_overlap() couldn't
find the overlap. Finally assign a wrong address to bar.
This patch using aligned new base to find the next rmrr, could fix the
above case and find all the overlapped rmrr with new base.
Jan Beulich [Tue, 12 Sep 2017 13:11:07 +0000 (15:11 +0200)]
gnttab: also validate PTE permissions upon destroy/replace
In order for PTE handling to match up with the reference counting done
by common code, presence and writability of grant mapping PTEs must
also be taken into account; validating just the frame number is not
enough. This is in particular relevant if a guest fiddles with grant
PTEs via non-grant hypercalls.
Note that the flags being passed to replace_grant_host_mapping()
already happen to be those of the existing mapping, so no new function
parameter is needed.
This is CVE-2017-14319 / XSA-234.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 16b1414de91b5a82a0996c67f6db3af7d7e32873
master date: 2017-09-12 14:45:13 +0200
tools/xenstore: dont unlink connection object twice
A connection object of a domain with associated stubdom has two
parents: the domain and the stubdom. When cleaning up the list of
active domains in domain_cleanup() make sure not to unlink the
connection twice from the same domain. This could happen when the
domain and its stubdom are being destroyed at the same time leading
to the domain loop being entered twice.
Additionally don't use talloc_free() in this case as it will remove
a random parent link, leading eventually to a memory leak. Use
talloc_unlink() instead specifying the context from which the
connection object should be removed.
This is CVE-2017-14317 / XSA-233.
Reported-by: Eric Chanudet <chanudete@ainfosec.com> Signed-off-by: Juergen Gross <jgross@suse.com> Reviewed-by: Ian Jackson <ian.jackson@eu.citrix.com>
master commit: 562a1c0f7ef3fbf3c122c3dfa4f2ad9dd51da9fe
master date: 2017-09-12 14:44:56 +0200
Andrew Cooper [Tue, 12 Sep 2017 13:10:11 +0000 (15:10 +0200)]
grant_table: fix GNTTABOP_cache_flush handling
Don't fall over a NULL grant_table pointer when the owner of the domain
is a system domain (DOMID_{XEN,IO} etc).
This is CVE-2017-14318 / XSA-232.
Reported-by: Matthew Daley <mattd@bugfuzz.com> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: c3d830b244998b3686e2eb64db95996be5eb5e5c
master date: 2017-09-12 14:44:11 +0200
George Dunlap [Tue, 12 Sep 2017 13:09:28 +0000 (15:09 +0200)]
xen/mm: make sure node is less than MAX_NUMNODES
The output of MEMF_get_node(memflags) can be as large as nodeid_t can
hold (currently 255). This is then used as an index to arrays of size
MAX_NUMNODE, which is 64 on x86 and 1 on ARM, can be passed in by an
untrusted guest (via memory_exchange and increase_reservation) and is
not currently bounds-checked.
Check the value in page_alloc.c before using it, and also check the
value in the hypercall call sites and return -EINVAL if appropriate.
Don't permit domains other than the hardware or control domain to
allocate node-constrained memory.
This is CVE-2017-14316 / XSA-231.
Reported-by: Matthew Daley <mattd@bugfuzz.com> Signed-off-by: George Dunlap <george.dunlap@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 2fece35303529395bfea6b03d2268380ef682c93
master date: 2017-09-12 14:43:16 +0200
When no memory is available in the hypervisor, rather than immediately
failing the request, try to steal a handle from another vCPU.
Reported-by: George Dunlap <george.dunlap@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: d02f1a0b7576bafb2fba903c7e6e7221ab0d2847
master date: 2017-08-17 14:41:01 +0200
cpufreq: only stop ondemand governor if already started
On CPUFREQ_GOV_STOP in cpufreq_governor_dbs, shortcut to
return success if the governor is already stopped.
Avoid executing dbs_timer_exit, to prevent tripping an assertion
within a call to kill_timer on a timer that has not been prepared
with init_timer, if the CPUFREQ_GOV_START case has not
run beforehand.
kill_timer validates timer state:
* itself, via BUG_ON(this_cpu(timers).running == timer);
* within active_timer, ASSERTing timer->status is within bounds;
* within list_del, which ASSERTs timer inactive list membership.
Patch is synonymous to an OpenXT patch produced at Citrix prior to
June 2014.
Signed-off-by: Christopher Clark <christopher.clark6@baesystems.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: e7ec6f5f32cd2d0f723083cde3d7761c4e675f2c
master date: 2017-08-10 12:35:50 +0200
Chao Gao [Mon, 28 Aug 2017 11:02:53 +0000 (13:02 +0200)]
VT-d PI: disable VT-d PI when CPU-side PI isn't enabled
From the context calling pi_desc_init(), we can conclude the current
implementation of VT-d PI depends on CPU-side PI. If we enable VT-d PI
and disable CPU-side PI by disabling APICv explicitly in xen boot
command line, we would get an assertion failure.
This patch clears iommu_intpost once finding CPU-side PI won't be enabled.
It is safe for this is done before this flag starts taking effect. Also
take this chance to remove the useless check of "acknowledge interrupt on
exit", which is a minimal requirement which has been checked earlier.
Signed-off-by: Chao Gao <chao.gao@intel.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Acked-by: Kevin Tian <kevin.tian@intel.com>
master commit: e489eb6138e7efe4214a7e9ba0d21f54fc5b7d35
master date: 2017-08-10 12:32:16 +0200
Rusty Bird [Mon, 28 Aug 2017 11:02:24 +0000 (13:02 +0200)]
VT-d: don't panic/warn on iommu=no-igfx
When operating on an Intel graphics device, iommu_enable_translation()
panicked (force_iommu==1) or warned (force_iommu==0) about the BIOS if
is_igd_vt_enabled_quirk() returned 0. That's good if the actual BIOS
problem has been detected. But since commit 1463411, returning 0 could
also happen if the user simply passed "iommu=no-igfx", in which case
bailing out with an info message (instead of a panic/warning) would be
more appropriate.
The panic broke the combination "iommu=force,no-igfx", and also the case
where "iommu=no-igfx" is passed but force_iommu=1 is set automatically
by x2apic_bsp_setup().
Move the iommu_igfx check from is_igd_vt_enabled_quirk() into its only
caller iommu_enable_translation(), and tweak the logic.
Olaf Hering [Mon, 28 Aug 2017 11:02:06 +0000 (13:02 +0200)]
docs: replace xm with xl in xen-tscmode
Signed-off-by: Olaf Hering <olaf@aepfle.de> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
master commit: 763267e315a93e2b6d66a0afdcda96db939e09b6
master date: 2017-07-24 10:17:21 +0100
Olaf Hering [Mon, 28 Aug 2017 11:01:50 +0000 (13:01 +0200)]
rombios: prevent building with PIC/PIE
If the default compiler silently defaults to to -fPIC/-fPIE building
rombios fails:
ld -melf_i386 -s -r 32bitbios.o tcgbios/tcgbiosext.o util.o pmm.o -o 32bitbios_all.o
There are undefined symbols in the BIOS:
U _GLOBAL_OFFSET_TABLE_
make[10]: *** [Makefile:26: 32bitbios_all.o] Error 11
Prevent the failure by enforcing non-PIC/PIE mode.
Signed-off-by: Olaf Hering <olaf@aepfle.de> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 672949d6c61d9cba01c5b414eed9d522082f04d3
master date: 2017-06-26 14:32:46 +0100
Andrew Cooper [Mon, 28 Aug 2017 11:01:23 +0000 (13:01 +0200)]
xen/livepatch: Don't crash on encountering STN_UNDEF relocations
A symndx of STN_UNDEF is special, and means a symbol value of 0. While
legitimate in the ELF standard, its existance in a livepatch is questionable
at best. Until a plausible usecase presents itself, reject such a relocation
with -EOPNOTSUPP.
Additionally, fix an off-by-one error while range checking symndx, and perform
a safety check on elf->sym[symndx].sym before derefencing it, to avoid
tripping over a NULL pointer when calculating val.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Stefano Stabellini <sstabellini@kernel.org> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Tested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> [x86 and arm32] Reviewed-by: Jan Beulich <JBeulich@suse.com> Reviewed-by: Ross Lagerwall <ross.lagerwall@citrix.com>
master commit: 2ff229643b739e2fd0cd0536ee9fca506cfa92f8
master date: 2017-06-23 15:00:37 +0100
Andrew Cooper [Mon, 28 Aug 2017 11:00:59 +0000 (13:00 +0200)]
xen/livepatch: Use zeroed memory allocations for arrays
Each of these arrays is sparse. Use zeroed allocations to cause uninitialised
array elements to contain deterministic values, most importantly for the
embedded pointers.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Tested-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> [x86 and arm32] Reviewed-by: Ross Lagerwall <ross.lagerwall@citrix.com>
master commit: bd53b85156aaf0330181ab9b82d9a6c52fb30f8c
master date: 2017-06-23 15:00:37 +0100
Andrew Cooper [Mon, 28 Aug 2017 11:00:07 +0000 (13:00 +0200)]
x86/hvm: Fixes to hvmemul_insn_fetch()
Force insn_off to a single byte, as offset can wrap around or truncate with
respect to sh_ctxt->insn_buf_eip under a number of normal circumstances.
Furthermore, don't use an ASSERT() for bounds checking the write into
hvmemul_ctxt->insn_buf[].
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Paul Durrant <paul.durrant@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
x86/hvm: Fix boundary check in hvmemul_insn_fetch()
c/s 0943a03037 added some extra protection for overflowing the emulation
instruction cache, but Coverity points out that boundary condition is off by
one when memcpy()'ing out of the buffer.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Paul Durrant <paul.durrant@citrix.com>
x86/HVM: fix boundary check in hvmemul_insn_fetch() (again)
Commit 5a992b670b ("x86/hvm: Fix boundary check in
hvmemul_insn_fetch()") went a little too far in its correction to
commit 0943a03037 ("x86/hvm: Fixes to hvmemul_insn_fetch()"): Keep the
start offset check, but restore the original end offset one.
Jan Beulich [Wed, 23 Aug 2017 15:54:33 +0000 (17:54 +0200)]
arm/mm: release grant lock on xenmem_add_to_physmap_one() error paths
Commit 55021ff9ab ("xen/arm: add_to_physmap_one: Avoid to map mfn 0 if
an error occurs") introduced error paths not releasing the grant table
lock. Replace them by a suitable check after the lock was dropped.
This is XSA-235.
Reported-by: Wei Liu <wei.liu2@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Julien Grall <julien.grall@arm.com>
master commit: 59546c1897a90fe9af5ebbbb05ead8d98b4d17b9
master date: 2017-08-23 17:45:45 +0200
Jan Beulich [Thu, 17 Aug 2017 13:11:22 +0000 (15:11 +0200)]
gnttab: fix transitive grant handling
Processing of transitive grants must not use the fast path, or else
reference counting breaks due to the skipped recursive call to
__acquire_grant_for_copy() (its __release_grant_for_copy()
counterpart occurs independent of original pin count). Furthermore
after re-acquiring temporarily dropped locks we need to verify no grant
properties changed if the original pin count was non-zero; checking
just the pin counts is sufficient only for well-behaved guests. As a
result, __release_grant_for_copy() needs to mirror that new behavior.
Furthermore a __release_grant_for_copy() invocation was missing on the
retry path of __acquire_grant_for_copy(), and gnttab_set_version() also
needs to bail out upon encountering a transitive grant.
This is part of XSA-226.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: ad48fb963dbff02762d2db5396fa655ac0c432c7
master date: 2017-08-17 14:40:31 +0200
Jan Beulich [Thu, 17 Aug 2017 13:10:50 +0000 (15:10 +0200)]
gnttab: don't use possibly unbounded tail calls
There is no guarantee that the compiler would actually translate them
to branches instead of calls, so only ones with a known recursion limit
are okay:
- __release_grant_for_copy() can call itself only once, as
__acquire_grant_for_copy() won't permit use of multi-level transitive
grants,
- __acquire_grant_for_copy() is fine to call itself with the last
argument false, as that prevents further recursion,
- __acquire_grant_for_copy() must not call itself to recover from an
observed change to the active entry's pin count
This is part of XSA-226.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 999d2ccb7f73408aa22656e1ba2f98b077eaa1c2
master date: 2017-08-17 14:39:18 +0200
Jan Beulich [Tue, 15 Aug 2017 13:26:42 +0000 (15:26 +0200)]
gnttab: correct pin status fixup for copy
Regardless of copy operations only setting GNTPIN_hst*, GNTPIN_dev*
also need to be taken into account when deciding whether to clear
_GTF_{read,writ}ing. At least for consistency with code elsewhere the
read part better doesn't use any mask at all.
This is XSA-230.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: 6e2a4c73564ab907b732059adb317d6ca2d138a2
master date: 2017-08-15 15:08:03 +0200
Jan Beulich [Tue, 15 Aug 2017 13:26:08 +0000 (15:26 +0200)]
gnttab: split maptrack lock to make it fulfill its purpose again
The way the lock is currently being used in get_maptrack_handle(), it
protects only the maptrack limit: The function acts on current's list
only, so races on list accesses are impossible even without the lock.
Otoh list access races are possible between __get_maptrack_handle() and
put_maptrack_handle(), due to the invocation of the former for other
than current from steal_maptrack_handle(). Introduce a per-vCPU lock
for list accesses to become race free again. This lock will be
uncontended except when it becomes necessary to take the steal path,
i.e. in the common case there should be no meaningful performance
impact.
When in get_maptrack_handle adds a stolen entry to a fresh, empty,
freelist, we think that there is probably no concurrency. However,
this is not a fast path and adding the locking there makes the code
clearly correct.
Also, while we are here: the stolen maptrack_entry's tail pointer was
not properly set. Set it.
This is CVE-2017-12136 / XSA-228.
Reported-by: Ian Jackson <ian.jackson@eu.citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Ian Jackson <Ian.Jackson@eu.citrix.com>
master commit: 02cbeeb6207508b0f04a2c6181445c8eb3f1e117
master date: 2017-08-15 15:07:25 +0200
Andrew Cooper [Tue, 15 Aug 2017 13:23:45 +0000 (15:23 +0200)]
x86/grant: disallow misaligned PTEs
Pagetable entries must be aligned to function correctly. Disallow attempts
from the guest to have a grant PTE created at a misaligned address, which
would result in corruption of the L1 table with largely-guest-controlled
values.
This is CVE-2017-12137 / XSA-227.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: ce442926c2530da9376199dcc769436376ad2386
master date: 2017-08-15 15:06:45 +0200
Jan Beulich [Fri, 23 Jun 2017 13:17:07 +0000 (15:17 +0200)]
memory: don't suppress P2M update in populate_physmap()
Commit d18627583d ("memory: don't hand MFN info to translated guests")
wrongly added a null-handle check there - just like stated in its
description for memory_exchange(), the array is also an input for
populate_physmap() (and hence can't reasonably be null). I have no idea
how I've managed to overlook this.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: b964e3106d2cdaa11cc4524181ff14607d110ae4
master date: 2017-06-20 14:51:53 +0200
When determining Access Rights, Protection Keys only take effect when CR4.PKE
it set, and 4-level paging is active. All other circumstances (notibly, 32bit
PAE paging) skip the Protection Key control mechanism.
Therefore, we do not need to clear CR4.PKE behind the back of a guest which is
not using paging, as such a guest is necesserily running with EFER.LMA
disabled.
The {RD,WR}PKRU instructions are specified as being legal for use in any
operating mode, but only if CR4.PKE is set. By clearing CR4.PKE behind the
back of an unpaged guest, these instructions yield #UD despite the guest
correctly seeing PKE set if it reads CR4, and OSPKE being visible in CPUID.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Huaitong Han <huaitong.han@intel.com> Acked-by: Kevin Tian <kevin.tian@intel.com>
master commit: 224acdd04a9f6ffe44d2f716287cac74787899ec
master date: 2017-06-01 14:13:57 +0100
Julien Grall [Tue, 20 Jun 2017 14:21:42 +0000 (16:21 +0200)]
xen/arm: vgic: Sanitize target mask used to send SGI
The current function vgic_to_sgi does not sanitize the target mask and
may therefore get an invalid vCPU ID. This will result to an out of
bound access of d->vcpu[...] as there is no check whether the vCPU ID is
within the maximum supported by the guest.
This was introduced by commit ea37fd2111 "xen/arm: split vgic driver
into generic and vgic-v2 driver".
Jan Beulich [Tue, 20 Jun 2017 14:21:09 +0000 (16:21 +0200)]
gnttab: __gnttab_unmap_common_complete() is all-or-nothing
All failures have to be detected in __gnttab_unmap_common(), the
completion function must not skip part of its processing. In particular
the GNTMAP_device_map related putting of page references and adjustment
of pin count must not occur if __gnttab_unmap_common() signaled an
error. Furthermore the function must not make adjustments to global
state (here: clearing GNTTAB_device_map) before all possibly failing
operations have been performed.
There's one exception for IOMMU related failures: As IOMMU manipulation
occurs after GNTMAP_*_map have been cleared already, the related page
reference and pin count adjustments need to be done nevertheless. A
fundamental requirement for the correctness of this is that
iommu_{,un}map_page() crash any affected DomU in case of failure.
The version check appears to be pointless (or could perhaps be a
BUG_ON() or ASSERT()), but for the moment also move it.
George Dunlap [Tue, 20 Jun 2017 14:20:41 +0000 (16:20 +0200)]
gnttab: correct logic to get page references during map requests
The rules for reference counting are somewhat complicated:
* Each of GNTTAB_host_map and GNTTAB_device_map need their own
reference count
* If the mapping is writeable:
- GNTTAB_host_map needs a type count under only some conditions
- GNTTAB_device_map always needs a type count
If the mapping succeeds, we need to keep all of these; if the mapping
fails, we need to release whatever references we have acquired so far.
Additionally, the code that does a lot of this calculation "inherits"
a reference as part of the process of finding out who the owner is.
Finally, if the grant is mapped as writeable (without the
GNTMAP_readonly flag), but the hypervisor cannot grab a
PGT_writeable_page type, the entire operation should fail.
Unfortunately, the current code has several logic holes:
* If a grant is mapped only GNTTAB_device_map, and with a writeable
mapping, but in conditions where a *host* type count is not
necessary, the code will fail to grab the necessary type count.
* If a grant is mapped both GNTTAB_device_map and GNTTAB_host_map,
with a writeable mapping, in conditions where the host type count is
not necessary, *and* where the page cannot be changed to type
PGT_writeable, the condition will not be detected.
In both cases, this means that on success, the type count will be
erroneously reduced when the grant is unmapped. In the second case,
the type count will be erroneously reduced on the failure path as
well. (In the first case the failure path logic has the same hole
as the reference grabbing logic.)
Additionally, the return value of get_page() is not checked; but this
may fail even if the first get_page() succeeded due to a reference
counting overflow.
First of all, simplify the restoration logic by explicitly counting
the reference and type references acquired.
Consider each mapping type separately, explicitly marking the
'incoming' reference as used so we know when we need to grab a second
one.
Finally, always check the return value of get_page[_type]() and go to
the failure path if appropriate.
This is part of XSA-224.
Reported-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: George Dunlap <george.dunlap@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com>
master commit: 75b384ece635adc55c2bafbdc2d8959c10542c31
master date: 2017-06-20 14:46:21 +0200
Jan Beulich [Tue, 20 Jun 2017 14:20:17 +0000 (16:20 +0200)]
gnttab: never create host mapping unless asked to
We shouldn't create a host mapping unless asked to even in the case of
mapping a granted MMIO page. In particular the mapping wouldn't be torn
down when processing the matching unmap request.
George Dunlap [Tue, 20 Jun 2017 14:19:41 +0000 (16:19 +0200)]
gnttab: fix handling of dev_bus_addr during unmap
If a grant has been mapped with the GNTTAB_device_map flag, calling
grant_unmap_ref() with dev_bus_addr set to zero should cause the
GNTTAB_device_map part of the mapping to be left alone.
Unfortunately, at the moment, op->dev_bus_addr is implicitly checked
before clearing the map and adjusting the pin count, but only the bits
above 12; and it is not checked at all before dropping page
references. This means a guest can repeatedly make such a call to
cause the reference count to drop to zero, causing the page to be
freed and re-used, even though it's still mapped in its pagetables.
To fix this, always check op->dev_bus_addr explicitly for being
non-zero, as well as op->flag & GNTMAP_device_map, before doing
operations on the device_map.
While we're here, make the logic a bit cleaner:
* Always initialize op->frame to zero and set it from act->frame, to reduce the
chance of untrusted input being used
* Explicitly check the full dev_bus_addr against act->frame <<
PAGE_SHIFT, rather than ignoring the lower 12 bits
This is part of XSA-224.
Reported-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: George Dunlap <george.dunlap@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com>
master commit: 8fdfcb2b6bcd074776560e76843815f124d587f1
master date: 2017-06-20 14:45:33 +0200
Julien Grall [Tue, 20 Jun 2017 14:19:24 +0000 (16:19 +0200)]
arm: vgic: Don't update the LR when the IRQ is not enabled
gic_raise_inflight_irq will be called if the IRQ is already inflight
(i.e the IRQ is injected to the guest). If the IRQ is already already in
the LRs, then the associated LR will be updated.
To know if the interrupt is already in the LR, the function check if the
interrupt is queued. However, if the interrupt is not enabled then the
interrupt may not be queued nor in the LR. So gic_update_one_lr may be
called (if we inject on the current vCPU) and read the LR.
Because the interrupt is not in the LR, Xen will either read:
* LR 0 if the interrupt was never injected before
* LR 255 (GIC_INVALID_LR) if the interrupt was injected once. This
is because gic_update_one_lr will reset p->lr.
Reading LR 0 will result to potentially update the wrong interrupt and
not keep the LRs in sync with Xen.
Reading LR 255 will result to:
* Crash Xen on GICv3 as the LR index is bigger than supported (see
gicv3_ich_read_lr).
* Read/write always GICH_LR + 255 * 4 that is not part of the memory
mapped.
The problem can be prevented by checking whether the interrupt is
enabled in gic_raise_inflight_irq before calling gic_update_one_lr.
A follow-up of this patch is expected to mitigate the issue in the
future.
Jan Beulich [Tue, 20 Jun 2017 14:18:43 +0000 (16:18 +0200)]
guest_physmap_remove_page() needs its return value checked
Callers, namely such subsequently freeing the page, must not blindly
assume success - the function may namely fail when needing to shatter a
super page, but there not being memory available for the then needed
intermediate page table.
As it happens, guest_remove_page() callers now also all check the
return value.
Furthermore a missed put_gfn() on an error path in gnttab_transfer() is
also being taken care of.
This is part of XSA-222.
Reported-by: Julien Grall <julien.grall@arm.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Julien Grall <julien.grall@arm.com> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: a0cce6048d010a30ac82f8db7787bbf9aada64f4
master date: 2017-06-20 14:41:16 +0200
Andrew Cooper [Tue, 20 Jun 2017 14:17:34 +0000 (16:17 +0200)]
memory: fix return value handing of guest_remove_page()
Despite the description in mm.h, guest_remove_page() previously returned 0 for
paging errors.
Switch guest_remove_page() to having regular 0/-error semantics, and propagate
the return values from clear_mmio_p2m_entry() and mem_sharing_unshare_page()
to the callers (although decrease_reservation() is the only caller which
currently cares).
This is part of XSA-222.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: b614f642c35da5184416787352f51a6379a92628
master date: 2017-06-20 14:39:56 +0200
Jan Beulich [Tue, 20 Jun 2017 14:17:09 +0000 (16:17 +0200)]
evtchn: avoid NULL derefs
Commit fbbd5009e6 ("evtchn: refactor low-level event channel port ops")
added a de-reference of the struct evtchn pointer for a port without
first making sure the bucket pointer is non-NULL. This de-reference is
actually entirely unnecessary, as all relevant callers (beyond the
problematic do_poll()) already hold the port number in their hands, and
the actual leaf functions need nothing else.
For FIFO event channels there's a second problem in that the ordering
of reads and updates to ->num_evtchns and ->event_array[] was so far
undefined (the read side isn't always holding the domain's event lock).
Add respective barriers.
Jan Beulich [Tue, 20 Jun 2017 14:16:24 +0000 (16:16 +0200)]
x86: avoid leaking PKRU and BND* between vCPU-s
PKRU is explicitly "XSAVE-managed but not XSAVE-enabled", so guests
might access the register (via {RD,WR}PKRU) without setting XCR0.PKRU.
Force context switching as well as migrating the register as soon as
CR4.PKE is being set the first time.
For MPX (BND<n>, BNDCFGU, and BNDSTATUS) the situation is less clear,
and the SDM has not entirely consistent information for that case.
While experimentally the instructions don't change register state as
long as the two XCR0 bits aren't both 1, be on the safe side and enable
both if BNDCFGS.EN is being set the first time.
This is XSA-220.
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
master commit: de20bb6c4f65c4161e0931402613f9ffac86302d
master date: 2017-06-20 14:36:51 +0200
Andrew Cooper [Tue, 20 Jun 2017 14:15:56 +0000 (16:15 +0200)]
x86/shadow: hold references for the duration of emulated writes
The (misnamed) emulate_gva_to_mfn() function translates a linear address to an
mfn, but releases its page reference before returning the mfn to its caller.
sh_emulate_map_dest() uses the results of one or two translations to construct
a virtual mapping to the underlying frames, completes an emulated
write/cmpxchg, then unmaps the virtual mappings.
The page references need holding until the mappings are unmapped, or the
frames can change ownership before the writes occurs.
This is XSA-219.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Tim Deegan <tim@xen.org>
master commit: 26217aff67ae1538d4e1b2226afab6993cdbe772
master date: 2017-06-20 14:36:11 +0200
Jan Beulich [Tue, 20 Jun 2017 14:15:24 +0000 (16:15 +0200)]
gnttab: correct maptrack table accesses
In order to observe a consistent (limit,pointer-table) pair, the reader
needs to either hold the maptrack lock (in line with documentation) or
both sides need to order their accesses suitably (the writer side
barrier was removed by commit dff515dfea ["gnttab: use per-VCPU
maptrack free lists"], and a read side barrier has never been there).
Make the writer publish a new table page before limit (for bounds
checks to work), and new list head last (for racing maptrack_entry()
invocations to work). At the same time add read barriers to lockless
readers.
Additionally get_maptrack_handle() must not assume ->maptrack_head to
not change behind its back: Another handle may be put (updating only
->maptrack_tail) and then got or stolen (updating ->maptrack_head).
This is part of XSA-218.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
master commit: 4b78efa91c8ae3c42e14b8eaeaad773c5eb3b71a
master date: 2017-06-20 14:34:34 +0200
George Dunlap [Tue, 20 Jun 2017 14:14:57 +0000 (16:14 +0200)]
gnttab: Avoid potential double-put of maptrack entry
Each grant mapping for a particular domain is tracked by an in-Xen
"maptrack" entry. This entry is is referenced by a "handle", which is
given to the guest when it calls gnttab_map_grant_ref().
There are two types of mapping a particular handle can refer to:
GNTMAP_host_map and GNTMAP_device_map. A given
gnttab_unmap_grant_ref() call can remove either only one or both of
these entries. When a particular handle has no entries left, it must
be freed.
gnttab_unmap_grant_ref() loops through its grant unmap request list
twice. It first removes entries from any host pagetables and (if
appropraite) iommus; then it does a single domain TLB flush; then it
does the clean-up, including telling the granter that entries are no
longer being used (if appropriate).
At the moment, it's during the first pass that the maptrack flags are
cleared, but the second pass that the maptrack entry is freed.
Unfortunately this allows the following race, which results in a
double-free:
A: (pass 1) clear host_map
B: (pass 1) clear device_map
A: (pass 2) See that maptrack entry has no mappings, free it
B: (pass 2) See that maptrack entry has no mappings, free it #
Unfortunately, unlike the active entry pinning update, we can't simply
move the maptrack flag changes to the second half, because the
maptrack flags are used to determine if iommu entries need to be
added: a domain's iommu must never have fewer permissions than the
maptrack flags indicate, or a subsequent map_grant_ref() might fail to
add the necessary iommu entries.
Instead, free the maptrack entry in the first pass if there are no
further mappings.
This is part of XSA-218.
Reported-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: George Dunlap <george.dunlap@citrix.com> Signed-off-by: Jan Beulich <jbeulich@suse.com>
master commit: b7f6cbb9d43f7384e1f38f8764b9a48216c8a525
master date: 2017-06-20 14:33:13 +0200
Jan Beulich [Tue, 20 Jun 2017 14:14:34 +0000 (16:14 +0200)]
gnttab: fix unmap pin accounting race
Once all {writable} mappings of a grant entry have been unmapped, the
hypervisor informs the guest that the grant entry has been released by
clearing the _GTF_{reading,writing} usage flags in the guest's grant
table as appropriate.
Unfortunately, at the moment, the code that updates the accounting
happens in a different critical section than the one which updates the
usage flags; this means that under the right circumstances, there may be
a window in time after the hypervisor reported the grant as being free
during which the grant referee still had access to the page.
Move the grant accounting code into the same critical section as the
reporting code to make sure this kind of race can't happen.
Quan Xu [Tue, 20 Jun 2017 14:14:06 +0000 (16:14 +0200)]
IOMMU: handle IOMMU mapping and unmapping failures
Treat IOMMU mapping and unmapping failures as a fatal to the DomU
If IOMMU mapping and unmapping failed, crash the DomU and propagate
the error up to the call trees.
No spamming of the log can occur. For DomU, we avoid logging any
message for already dying domains. For Dom0, that'll still be more
verbose than we'd really like, but it at least wouldn't outright
flood the console.
Signed-off-by: Quan Xu <quan.xu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: 834c97baebb3743c54bcae228e984ae1b9692e6a
master date: 2016-06-14 15:10:57 +0200
Jan Beulich [Tue, 20 Jun 2017 14:13:15 +0000 (16:13 +0200)]
x86/mm: disallow page stealing from HVM domains
The operation's success can't be controlled by the guest, as the device
model may have an active mapping of the page. If we nevertheless
permitted this operation, we'd have to add further TLB flushing to
prevent scenarios like
"Domains A (HVM), B (PV), C (PV); B->target==A
Steps:
1. B maps page X from A as writable
2. B unmaps page X without a TLB flush
3. A sends page X to C via GNTTABOP_transfer
4. C maps page X as pagetable (potentially causing a TLB flush in C,
but not in B)
At this point, X would be mapped as a pagetable in C while being
writable through a stale TLB entry in B."
A similar scenario could be constructed for A using XENMEM_exchange and
some arbitrary PV domain C then having this page allocated.
This is XSA-217.
Reported-by: Jann Horn <jannh@google.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: George Dunlap <george.dunlap@citrix.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
master commit: fae7d5be8bb8b7a5b7005c4f3b812a47661a721e
master date: 2017-06-20 14:29:51 +0200
vgic: refuse irq migration when one is already in progress
When an irq migration is already in progress, but not yet completed
(GIC_IRQ_GUEST_MIGRATING is set), refuse any other irq migration
requests for the same irq.
This patch implements this approach by returning success or failure from
vgic_migrate_irq, and avoiding irq target changes on failure. It prints
a warning in case the irq migration fails.
It also moves the clear_bit of GIC_IRQ_GUEST_MIGRATING to after the
physical irq affinity has been changed so that all operations regarding
irq migration are completed.
arm: remove irq from inflight, then change physical affinity
This patch fixes a potential race that could happen when
gic_update_one_lr and vgic_vcpu_inject_irq run simultaneously.
When GIC_IRQ_GUEST_MIGRATING is set, we must make sure that the irq has
been removed from inflight before changing physical affinity, to avoid
concurrent accesses to p->inflight, as vgic_vcpu_inject_irq will take a
different vcpu lock.
Julien Grall [Fri, 5 May 2017 14:30:36 +0000 (15:30 +0100)]
xen/arm: Survive unknown traps from guests
Currently we crash Xen if we see an ESR_EL2.EC value we don't recognise.
As configurable disables/enables are added to the architecture
(controlled by RES1/RESO bits respectively), with associated synchronous
exceptions, it may be possible for a guest to trigger exceptions with
classes that we don't recognise.
While we can't service these exceptions in a manner useful to the guest,
we can avoid bringing down the host. Per ARM DDI 0487A.k_iss10775, page
D7-1937, EC values within the range 0x00 - 0x2c are reserved for future
use with synchronous exceptions, and EC within the range 0x2d - 0x3f may
be used for either synchronous or asynchronous exceptions.
The patch makes Xen handle any unknown EC by injecting an UNDEFINED
exception into the guest, with a corresponding (ratelimited) warning in
the log.
This patch is based on Linux commit f050fe7a9164 "arm: KVM: Survive unknown
traps from the guest".
Julien Grall [Fri, 5 May 2017 14:30:35 +0000 (15:30 +0100)]
xen/arm: do_trap_hypervisor: Separate hypervisor and guest traps
The function do_trap_hypervisor is currently handling both trap coming
from the hypervisor and the guest. This makes difficult to get specific
behavior when a trap is coming from either the guest or the hypervisor.
Split the function into two parts:
- do_trap_guest_sync to handle guest traps
- do_trap_hyp_sync to handle hypervisor traps
On AArch32, the Hyp Trap Exception provides the standard mechanism for
trapping Guest OS functions to the hypervisor (see B1.14.1 in ARM DDI
0406C.c). It cannot be generated when generated when the processor is in
Hyp Mode, instead other exception will be used. So it is fine to replace
the call to do_trap_hypervisor by do_trap_guest_sync.
For AArch64, there are two distincts exception depending whether the
exception was taken from the current level (hypervisor) or lower level
(guest).
Note that the unknown traps from guests will lead to panic Xen. This is
already behavior and is left unchanged for simplicy. A follow-up patch
will address that.
xen/arm: Save ESR_EL2 to avoid using mismatched value in syndrome check
Xen will do exception syndrome check while some types of exception
take place in EL2. The syndrome check code read the ESR_EL2 register
directly, but in some situation this register maybe overridden by
nested exception.
For example, if we re-enable IRQ before reading ESR_EL2 which means
Xen may enter in IRQ exception mode and return the processor with
clobbered ESR_EL2 (See ARM ARM DDI 0487A.j D7.2.25)
In this case the guest exception syndrome has been overridden, we will
check the syndrome for guest sync exception with an incorrect ESR_EL2
value. So we want to save ESR_EL2 to cpu_user_regs as soon as the
exception takes place in EL2 to avoid using an incorrect syndrome value.
In order to save ESR_EL2, we added a 32-bit member hsr to cpu_user_regs.
But while saving registers in trap entry, we use stp to save ELR and
CPSR at the same time through 64-bit general registers. If we keep this
code, the hsr will be overridden by upper 32-bit of CPSR. So adjust the
code to use str to save ELR in a separate instruction and use stp to
save CPSR and HSR at the same time through 32-bit general registers.
This change affects the registers restore in trap exit, we can't use the
ldp to restore ELR and CPSR from stack at the same time. We have to use
ldr to restore them separately.
projects / xen.git / log