He Chen [Mon, 5 Sep 2016 10:49:43 +0000 (12:49 +0200)]
x86: allow disabling sm{e,a}p for Xen itself
SMEP/SMAP is a security feature to prevent kernel executing/accessing
user address involuntarily, any such behavior will lead to a page fault.
SMEP/SMAP is open (in CR4) for both Xen and HVM guest in earlier code.
SMEP/SMAP bit set in Xen CR4 would enforce security checking for 32-bit
PV guest which will suffer unknown SMEP/SMAP page fault when guest
kernel attempt to access user address although SMEP/SMAP is close for
PV guests.
This patch introduces a new boot option value "hvm" for "sm{e,a}p", it
is going to diable SMEP/SMAP for Xen hypervisor while enable them for
HVM. In this way, 32-bit PV guest will not suffer SMEP/SMAP security
issue. Users can choose whether open SMEP/SMAP for Xen itself,
especially when they are going to run 32-bit PV guests.
Signed-off-by: He Chen <he.chen@linux.intel.com>
[jbeulich: doc and style adjustments] Reviewed-by: Jan Beulich <jbeulich@suse.com>
have __DEFINE_COMPAT_HANDLE() generate const versions
Both DEFINE_XEN_GUEST_HANDLE() and __DEFINE_XEN_GUEST_HANDLE()
each produce both const and non-const handles,
only DEFINE_COMPAT_HANDLE() does (__DEFINE_COMPAT_HANDLE()
does not). This patch has __DEFINE_COMPAT_HANDLE() also
produce a const handle.
Suggested-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com> Acked-by: Jan Beulich <jbeulich@suse.com>
Tamas K Lengyel [Mon, 5 Sep 2016 10:47:16 +0000 (12:47 +0200)]
x86/monitor: include EAX/ECX in CPUID monitor events
Extend the CPUID monitor event to include EAX and ECX values that were used
when CPUID was executed. This is useful in identifying which leaf was queried.
We also adjust the xen-access output format to more closely resemble the output
of the Linux cpuid tool's raw format.
Signed-off-by: Tamas K Lengyel <tamas.lengyel@zentific.com> Acked-by: Razvan Cojocaru <rcojocaru@bitdefender.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Luwei Kang [Mon, 5 Sep 2016 10:46:13 +0000 (12:46 +0200)]
x86/cpuid: AVX-512 feature detection
AVX512 is an extention of AVX2. Its spec can be found at:
https://software.intel.com/sites/default/files/managed/b4/3a/319433-024.pdf
This patch detects AVX512 features by CPUID.
Signed-off-by: Luwei Kang <luwei.kang@intel.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
Jan Beulich [Fri, 2 Sep 2016 12:22:28 +0000 (14:22 +0200)]
x86/mm: drop pointless use of __FUNCTION__
Non-debugging message text should be (and is here) distinguishable
without also logging function names.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: George Dunlap <george.dunlap@citrix.com>
Jan Beulich [Fri, 2 Sep 2016 12:19:51 +0000 (14:19 +0200)]
memory: fix compat handling of XENMEM_access_op
Within compat_memory_op() this needs to be placed in the first switch()
statement, or it ends up being dead code (as that first switch() has a
default case chaining to compat_arch_memory_op()).
Signed-off-by: Jan Beulich <jbeulich@suse.com> Tested-by: Razvan Cojocaru <rcojocaru@bitdefender.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Fri, 2 Sep 2016 12:19:29 +0000 (14:19 +0200)]
x86/PV: make PMU MSR handling consistent
So far accesses to Intel MSRs on an AMD system fall through to the
default case, while accesses to AMD MSRs on an Intel system bail (in
the RDMSR case without updating EAX and EDX). Make the "AMD MSRs on
Intel" case match the "Intel MSR on AMD" one.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Fri, 2 Sep 2016 12:18:52 +0000 (14:18 +0200)]
x86: correct PT_NOTE file position
Program and section headers disagreed about the file offset at which
the build ID note lives.
Reported-by: Sylvain Munaut <s.munaut@whatever-company.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
credit1: fix a race when picking initial pCPU for a vCPU
In the Credit1 hunk of 9f358ddd69463 ("xen: Have
schedulers revise initial placement") csched_cpu_pick()
is called without taking the runqueue lock of the
(temporary) pCPU that the vCPU has been assigned to
(e.g., in XEN_DOMCTL_max_vcpus).
However, although 'hidden' in the IS_RUNQ_IDLE() macro,
that function does access the runq (for doing load
balancing calculations). Two scenarios are possible:
1) we are on cpu X, and IS_RUNQ_IDLE() peeks at cpu's
X own runq;
2) we are on cpu X, but IS_RUNQ_IDLE() peeks at some
other cpu's runq.
Scenario 2) absolutely requies that the appropriate
runq lock is taken. Scenario 1) works even without
taking the cpu's own runq lock. That is actually what
happens when when _csched_pick_cpu() is called from
csched_vcpu_acct() (in turn, called by csched_tick()).
Races have been observed and reported (by both XenServer
own testing and OSSTest [1]), in the form of
IS_RUNQ_IDLE() falling over LIST_POISON, because we're
not currently holding the proper lock, in
csched_vcpu_insert(), when scenario 1) occurs.
However, for better robustness, from now on we always
ask for the proper runq lock to be held when calling
IS_RUNQ_IDLE() (which is also becoming a static inline
function instead of macro).
In order to comply with that, we take the lock around
the call to _csched_cpu_pick() in csched_vcpu_acct().
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com> Signed-off-by: Dario Faggioli <dario.faggioli@citrix.com> Reviewed-by: George Dunlap <george.dunlap@citrix.com>
Andrew Cooper [Sat, 2 Jul 2016 10:43:02 +0000 (11:43 +0100)]
xen/trace: Turn the stub debugtrace_{dump,printk}() macros into functions
This allows printf format checking to be performed, and for
debugtrace_printk() to evaluate its arguments, even if debugtrace is disabled
at compile time.
No intended change.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com>
Andrew Cooper [Sat, 2 Jul 2016 10:28:13 +0000 (11:28 +0100)]
x86/shadow: More consistent printing for debug messages
* Use %pv or just d%d in preference to the multiple current ways of
presenting the same information.
* Use PRI_mfn instead of opencoding it.
* Drop all explicit use of __func__ from SHADOW_{PRINTK,DEBUG}() calls. The
wrappers already include it.
* Use hex rather than decimal for printing a pagefault error code.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Tim Deegan <tim@xen.org>
Juergen Gross [Tue, 30 Aug 2016 14:53:39 +0000 (16:53 +0200)]
stubdom: support Mini-OS config for Mini-OS apps
Mini-OS apps need to be compiled with the appropriate config settings
of Mini-OS, as there are various dependencies on those settings in
header files included by the apps.
Enhance stubdom Makefile to set the appropriate CPPFLAGS when calling
the apps' make.
Signed-off-by: Juergen Gross <jgross@suse.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
[ wei: fold in change to Config.mk to update mini-os commit ]
Andrew Cooper [Thu, 1 Sep 2016 09:45:03 +0000 (10:45 +0100)]
tools/migrate: Prevent PTE truncation from being fatal duing the live phase
It is possible, when normalising a PV pagetable that the table has been freed
and reused for something else by the guest.
In such a case, data read might no longer be a pagetable, and fail the
truncation check. However, this should only be fatal if we encounter such a
page in the paused phase.
This check is now consistent with all other checks in the same area.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Andrew Cooper [Fri, 2 Sep 2016 06:12:29 +0000 (08:12 +0200)]
x86/levelling: fix breakage on older Intel boxes from c/s 08e7738
cpufeat_mask() yields an unsigned integer constant. As a result, taking its
complement causes zero extention rather than sign extention.
The result is that, when a guest OS has OXSAVE disabled, all features in 1d
are hidden from native CPUID. Amongst other things, this causes the early
code in Linux to find no LAPIC, but for everything to appear fine later when
userspace is up and running.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Tested-by: Jan Beulich <jbeulich@suse.com>
Jan Beulich [Thu, 1 Sep 2016 13:23:46 +0000 (15:23 +0200)]
x86/32on64: misc adjustments to call gate emulation
- There's no 32-bit displacement in 16-bit addressing mode.
- It is wrong to ASSERT() anything on parts of an instruction fetched
from guest memory.
- The two scaling bits of a SIB byte don't affect whether there is a
scaled index register or not.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Thu, 1 Sep 2016 13:21:06 +0000 (15:21 +0200)]
x86: drop pointless uses of __func__ / __FUNCTION__
Non-debugging message text should be (and is in the cases here)
distinguishable without also logging function names. Debugging message
text, otoh, already includes file name and line number, so also
logging function names is redundant. One relatively pointless debugging
message gets removed altogether. In another case a missing log level
specifier gets added at once.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Thu, 1 Sep 2016 13:19:40 +0000 (15:19 +0200)]
x86/EFI: use less crude a way of generating the build ID
Recent enough binutils (2.25 onwards) support --build-id also for
COFF/PE output, and hence we should use that in favor of the original
hack when possible.
This gets complicated by the linker requiring at least one COFF object
file to attach the .buildid section to. Hence the patch introduces a
buildid.ihex (in order to avoid introducing binary files into the repo)
which then gets converted to a binary minimal COFF object (no sections,
no symbols).
Also (to avoid both code fragment going out of sync) remove an unneeded
ALIGN() from xen.lds.S: Adding an equivalent of it to the .buildid
section would cause the _erodata symbol to become associated with the
wrong section again (see commit 0970299de5 ["x86/EFI + Live Patch:
avoid symbol address truncation"]). And it's pointless because the
alignment already gets properly set by the input section(s).
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Acked-by: Andrew Cooper <andrew.cooper3@citrix.com>
Andrew Cooper [Mon, 22 Aug 2016 16:50:55 +0000 (17:50 +0100)]
x86/levelling: Provide architectural OSXSAVE handling to masked native CPUID
Contrary to c/s b2507fe7 "x86/domctl: Update PV domain cpumasks when setting
cpuid policy", Intel CPUID masks are applied after fast forwarding hardware
state, rather than before. (All behaviour in this regard appears completely
undocumented by both Intel and AMD).
Therefore, a set bit in the MSR causes hardware to be fast-forwarded, while a
clear bit forces the guests view to 0, even if Xen's CR4.OSXSAVE is actually
set.
This allows Xen to provide an architectural view of a guest kernels
CR4.OSXSAVE setting to any native CPUID instruction issused by guest kernel or
userspace, even when masking is used.
The masking value defaults to 1 (if the guest has XSAVE available) to cause
fast-forwarding to occur for the HVM and idle vcpus.
When setting the MSRs, a PV guest kernel's choice of OXSAVE is taken into
account, and clobbered from the MSR if not set. This causes the
fast-forwarding of Xen's CR4 state not to happen.
As a side effect however, levelling potentially need updating on all PV CR4
changes.
Reported-by: Jan Beulich <JBeulich@suse.com> Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
Wei Liu [Wed, 31 Aug 2016 15:26:51 +0000 (16:26 +0100)]
xen: fix gcov compilation
Currently enabling gcov in hypervisor won't build because although 26c9d03d ("gcov: Adding support for coverage information") claimed that
%.init.o files were excluded from applying compilation options, it was
in fact not true.
Fix that by filtering out the options correctly. Because the dependency
of stub.o in x86 EFI build can't be eliminated easily and we prefer a
generalised method going forward, we introduce nogcov-y to explicitly
mark objects that don't need to build with gcov support.
Signed-off-by: Wei Liu <wei.liu2@citrix.com> Signed-off-by: Jan Beulich <JBeulich@suse.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Wei Liu [Wed, 31 Aug 2016 15:26:50 +0000 (16:26 +0100)]
arm64: use "b" to branch to start_xen
The cbz instruction has range limitation. When compiled with gcov
support the object is larger so cbz can't handle that anymore. The error
message is like:
aarch64-linux-gnu-ld -EL -T xen.lds -N prelink.o \
/local/work/xen.git/xen/common/symbols-dummy.o -o /local/work/xen.git/xen/.xen-syms.0
prelink.o: In function `launch':
/local/work/xen.git/xen/arch/arm/arm64/head.S:602:(.text+0x408): relocation truncated to fit: R_AARCH64_CONDBR19 against symbol `start_xen' defined in .init.text section in prelink.o
Use "b" instead.
Signed-off-by: Wei Liu <wei.liu2@citrix.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Reviewed-by: Julien Grall <julien.grall@arm.com>
Wei Liu [Fri, 26 Aug 2016 10:11:46 +0000 (11:11 +0100)]
libxl: update flex output files
Libxl ships output files from flex (libxlu_*_l.{c,h}). We use the flex
shipped in Debian to generate those files. Debian just patched their
flex (DSA 3653-1) to fix CVE-2016-6354, which is a buffer overrun bug.
Note that libxl is _NOT_ vulnerable to that CVE. See below for Ian's
analysis to security@xen.
It would still be nice that we update our shipped flex output files to
avoid confusion.
===QUOTE===
The bug is that with input >16K[1] flex would usually fail to resize
the input buffer, and then overrun it.
I have read the code in libxlu_cfg_l.c to try to understand the
implications for libxl.
AFAICT
- libxl always does config file reading _from the file_ itself, and
provides flex with a string or buffer.
- so we always call whatever_yy_scan_bytes, not any other flex setup
function to set up a `buffer' (as flex calls it)
- yy_scan_bytes calls yy_scan_buffer to set up the buffer
- yy_scan_buffer sets b->yy_fill_buffer
- The effect of this is that yy_get_next_buffer will always
return early, rather than continuing on to the vulnerable code.
So I think libxl is not vulnerable, regardless of the contents of the
configuration file.
[1] the default buffer size, or whatever other buffer size is
configured (but we don't change it)
===ENDQUOTE===
Signed-off-by: Wei Liu <wei.liu2@citrix.com> Acked-by: Ian Jackson <ian.jackson@eu.citrix.com>
Juergen Gross [Fri, 26 Aug 2016 11:58:55 +0000 (13:58 +0200)]
libxc: correct max_pfn calculation for saving domain
Commit 91e204d37f44913913776d0a89279721694f8b32 ("libxc: try to find
last used pfn when migrating") introduced a bug for the case of a
domain supporting the virtual mapped linear p2m list: the maximum pfn
of the domain calculated from the p2m memory allocation might be too
low.
Correct this.
Reported-by: Stefan Bader <stefan.bader@canonical.com> Signed-off-by: Juergen Gross <jgross@suse.com> Tested-by: Stefan Bader <stefan.bader@canonical.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
The struct hvm_domain.vmx is defined in a union along with the svm.
This can causes issue for SVM since this code is used in the common
scheduling code for x86. The logic must check for cpu_has_vmx before
accessing the hvm_domain.vmx sturcture.
Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> Acked-by: Jan Beulich <jbeulich@suse.com>
Jan Beulich [Mon, 29 Aug 2016 14:04:22 +0000 (16:04 +0200)]
pass-through: drop pointless uses of __func__
Non-debugging message text should be (and is in the cases here)
distinguishable without also logging function names. Additionally log
the PCI device coordinates for alloc_pdev() failure.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Daniel Kiper [Thu, 25 Aug 2016 12:03:24 +0000 (14:03 +0200)]
x86/boot: use %ecx instead of %eax
Use %ecx instead of %eax to store low memory upper limit from EBDA.
This way we do not wipe multiboot protocol identifier. It is needed
in reloc() to differentiate between multiboot (v1) and
multiboot2 protocol.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Daniel Kiper [Thu, 25 Aug 2016 12:02:53 +0000 (14:02 +0200)]
x86/boot: call reloc() using stdcall calling convention
Current reloc() call method makes confusion and does not scale well
for more arguments. And subsequent patch adding multiboot2 protocol
support have to pass 3 arguments instead of 2. Hence, move reloc()
call to stdcall calling convention. One may argue that we should use
standard cdecl calling convention. However, stdcall is better here
than cdecl because we do not need to remove "manually" arguments from
stack in xen/arch/x86/boot/head.S assembly file.
Suggested-by: Jan Beulich <jbeulich@suse.com> Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com> Acked-by: Jan Beulich <jbeulich@suse.com>
Daniel Kiper [Thu, 25 Aug 2016 12:00:57 +0000 (14:00 +0200)]
x86/boot: create *.lnk files with linker script
Newer GCC (e.g. gcc version 5.1.1 20150618 (Red Hat 5.1.1-4) (GCC)) does
some code optimizations by creating data sections (e.g. jump addresses
for C switch/case are calculated using data in .rodata section). This
thing is not accepted by *.lnk build recipe which requires that only .text
section lives in output. Potentially we can inhibit this GCC behavior by
using special options, e.g. -fno-tree-switch-conversion. However, this
does not guarantee that in the future new similar optimizations or anything
else which creates not accepted sections will not break our build recipes
again. I do not mention that probably this is not good idea to just disable
random optimizations. So, take over full control on *.lnk linking process
by using linker script and merge all text and data sections into one
.text section.
Additionally, remove .got.plt section which is not used in our final code.
Signed-off-by: Daniel Kiper <daniel.kiper@oracle.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
Chris Patterson [Thu, 25 Aug 2016 07:00:59 +0000 (09:00 +0200)]
ns16550: mask transmit holding register empty interrupt when tx is stopped
The uart generates an interrupt whenever the transmit holding register is
empty and UART_IER_ETHREI is set in UART_IER. Currently, Xen's ns16550
driver does not currently mask this interrupt when transmit is stopped,
unlike other platforms such as Linux [1].
Toggle UART_IER_ETHREI flag in the UART_IER according to the state dictated
by stop_tx and start_tx hooks.
On the Tegra platform (forthcoming series), the reset via reading IIR does not
prevent re-assertion of THRE. This causes Xen to hang in the interrupt
handler's while loop whenever there is no data to transmit. This behavior (bug?)
is addressed by utilizing the start & stop tx hooks.
This has been tested on various x86 PCs for any obvious signs of regressions.
Wei Liu [Mon, 22 Aug 2016 12:47:53 +0000 (13:47 +0100)]
hvmloader: use bound checking in get_module_entry
Coverity complains:
overflow_before_widen: Potentially overflowing expression
info->nr_modules * 32U with type unsigned int (32 bits, unsigned) is
evaluated using 32-bit arithmetic, and then used in a context that
expects an expression of type uint64_t (64 bits, unsigned).
The overflow is unlikely to happen in reality because we only expect a
few modules.
Fix that by converting the check to use bound checking to placate
Coverity.
Signed-off-by: Wei Liu <wei.liu2@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com>
Andrew Cooper [Fri, 19 Aug 2016 14:08:10 +0000 (15:08 +0100)]
xen/physmap: Do not permit a guest to populate PoD pages for itself
PoD is supposed to be entirely transparent to guest, but this interface has
been left exposed for a long time.
The use of PoD requires careful co-ordination by the toolstack with the
XENMEM_{get,set}_pod_target hypercalls, and xenstore ballooning target. The
best a guest can do without toolstack cooperation crash.
Furthermore, there are combinations of features (e.g. c/s c63868ff "libxl:
disallow PCI device assignment for HVM guest when PoD is enabled") which a
toolstack might wish to explicitly prohibit (in this case, because the two
simply don't function in combination). In such cases, the guest mustn't be
able to subvert the configuration chosen by the toolstack.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com>
Jan Beulich [Fri, 19 Aug 2016 15:04:28 +0000 (17:04 +0200)]
x86: don't needlessly globalize page table labels
Neither l1_identmap[] nor l3_identmap[] get referenced from outside
their defining source file; the latter didn't even have an extern
declaration for use from C sources.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Fri, 19 Aug 2016 15:03:33 +0000 (17:03 +0200)]
x86/EFI: don't apply relocations to l{2,3}_bootmap
Other than claimed in commit 2ce5963727's ("x86: construct the
{l2,l3}_bootmap at compile time") the initialization of the two page
tables doesn't take care of everything without furher adjustment: The
compile time initialization obviously requires base relocations, and
those get processed after efi_arch_memory_setup(). Hence without
additional care the correctly initialized values may then get wrongly
"adjusted" again. Except the two table from being subject to base
relocation.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper@citrix.com>
Derek Straka [Fri, 19 Aug 2016 15:02:27 +0000 (17:02 +0200)]
x86: add a tboot Kconfig option
Allows for the conditional inclusion of tboot related functionality
via Kconfig
The default configuration for the new CONFIG_TBOOT option is 'y', so the
behavior out of the box remains unchanged. The addition of the option allows
advanced users to disable system behaviors associated with tboot at compile
time rather than relying on the run-time detection and configuration.
The CONFIG_CRYPTO option is 'n' by default and selected by the individual users
that require the functionality. Currently, the only user is tboot.
Signed-off-by: Derek Straka <derek@asterius.io> Acked-by: Jan Beulich <jbeulich@suse.com>
Anthony PERARD [Thu, 8 Oct 2015 15:26:32 +0000 (16:26 +0100)]
configure: do not depend on SEABIOS_PATH or OVMF_PATH ...
... to compile SeaBIOS and OVMF. Only depend on CONFIG_*.
If --with-system-* configure option is used, then set *_CONFIG=n to not
compile SEABIOS and OVMF.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
Anthony PERARD [Wed, 7 Oct 2015 14:45:14 +0000 (15:45 +0100)]
hvmloader: Load OVMF from modules
... and do not include the OVMF ROM into hvmloader anymore.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Anthony PERARD [Tue, 20 Oct 2015 15:57:51 +0000 (16:57 +0100)]
hvmloader: Load SeaBIOS from hvm_start_info modules
... and do not include the SeaBIOS ROM into hvmloader anymore.
This also fix the dependency on roms.inc, hvmloader.o does not include it.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Anthony PERARD [Mon, 19 Oct 2015 14:42:14 +0000 (15:42 +0100)]
hvmloader: Grab the hvm_start_info pointer
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Acked-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Anthony PERARD [Wed, 16 Sep 2015 14:12:15 +0000 (15:12 +0100)]
libxl: Load guest BIOS from file
The path to the BIOS blob can be overriden by the xl's
bios_path_override option, or provided by u.hvm.bios_firmware in the
domain_build_info struct by other libxl user.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
Anthony PERARD [Mon, 28 Sep 2015 18:03:55 +0000 (19:03 +0100)]
configure: #define SEABIOS_PATH and OVMF_PATH
Those paths are to be used by libxl, in order to load the firmware in
memory. If a system path is not defined via --with-system-seabios or
--with-system-ovmf, then default to the Xen firmware directory.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
Anthony PERARD [Mon, 21 Sep 2015 11:36:25 +0000 (12:36 +0100)]
libxc: Prepare a start info structure for hvmloader
... and load BIOS/UEFI firmware into guest memory.
This adds a new firmware module, system_firmware_module. It is loaded in
the guest memory and final location is provided to hvmloader via the
hvm_start_info struct.
This patch create the hvm_start_info struct for HVM guest that have a
device model, so this is now common code with HVM guest without device
model.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
Anthony PERARD [Fri, 19 Feb 2016 17:35:43 +0000 (17:35 +0000)]
libxc: Rework extra module initialisation
This patch use xc_dom_alloc_segment() to allocate the memory space for the
ACPI modules and the SMBIOS modules. This is to replace the arbitrary
placement of 1MB (+ extra for MB alignement) after the hvmloader image.
This patch can help if one add extra ACPI table and hvmloader contain
OVMF (OVMF is a 2MB binary), as in that case the extra ACPI table could
easily be loaded past the address 4MB, but hvmloader use a range of
memory from 4MB to 10MB to perform tests and in the process, clears the
memory, before loading the modules.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
Juergen Gross [Thu, 18 Aug 2016 10:04:30 +0000 (12:04 +0200)]
xen: Move the hvm_start_info C representation to the public headers
Instead of having several representation of hvm_start_info in C, define
it in public/arch-x86/hvm/start_info.h so both libxc and hvmloader can
use it.
Also move the comment describing the binary format to be alongside the
C struct.
Signed-off-by: Anthony PERARD <anthony.perard@citrix.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Jan Beulich <jbeulich@suse.com> Acked-by: Wei Liu <wei.liu2@citrix.com>
Jan Beulich [Wed, 17 Aug 2016 13:36:55 +0000 (15:36 +0200)]
x86emul: improve LOCK handling
Certain opcodes would so far not have got #UD when a LOCK prefix was
present. Adjust this by
- moving the too early generic check into destination operand decoding,
where DstNone and DstReg already have respective handling
- switching source and destination of TEST r,r/m, for it to be taken
care of by aforementioned generic checks
- explicitly dealing with all forms of CMP, SHLD, SHRD, as well as
TEST $imm,r/m
To make the handling of opcodes F6 and F7 more obvious, reduce the
amount of state set in the table, and adjust the respective switch()
statement accordingly.
Also eliminate the latent bug of the check in DstNone handling not
considering the opcode extension set.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Wed, 17 Aug 2016 13:34:26 +0000 (15:34 +0200)]
x86emul: introduce SrcEax for XCHG
Just like said in commit c0bc0adf24 ("x86emul: use DstEax where
possible"): While it avoids just a few instructions, we should
nevertheless make use of generic code as much as possible. Here we can
arrange for that by simply introducing SrcEax (which requires no other
code adjustments).
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Wed, 17 Aug 2016 13:33:27 +0000 (15:33 +0200)]
x86emul: don't open code EFLAGS handling for 2-operand IMUL
Slightly extending the emulate_2op*() macro machinery makes it usable
for IMUL r,r/m too, which has the benefit of smaller source code and
the EFLAGS output being guaranteed to match actual hardware behavior.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Wed, 17 Aug 2016 13:32:51 +0000 (15:32 +0200)]
x86emul: use DstEax also for {,I}{MUL,DIV}
Just like said in commit c0bc0adf24 ("x86emul: use DstEax where
possible"): While it avoids just a few instructions, we should
nevertheless make use of generic code as much as possible.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Wed, 17 Aug 2016 13:31:56 +0000 (15:31 +0200)]
domctl: relax getdomaininfo permissions
Qemu needs access to this for the domain it controls, both due to it
being used by xc_domain_memory_mapping() (which qemu calls) and the
explicit use in hw/xenpv/xen_domainbuild.c:xen_domain_poll(). Extend
permissions to that of any "ordinary" domctl: A domain controlling the
targeted domain can invoke this operation for that target domain (which
is being achieved by no longer passing NULL to xsm_domctl()).
This at once avoids a for_each_domain() loop when the ID of an
existing domain gets passed in.
Reported-by: Marek Marczykowski-Górecki <marmarek@invisiblethingslab.com> Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com> Acked-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Jan Beulich [Tue, 16 Aug 2016 13:44:05 +0000 (15:44 +0200)]
x86emul: drop SrcInvalid
As of commit a800e4f611 ("x86emul: drop pointless and add useful
default cases") we no longer need the early bailing when "d == 0" (the
default cases in the main switch() statements take care of that),
removal of which renders both callers of internal_error() wrong and
SrcInvalid useless. Drop them, as they're going to get in the way of
completing the decoder to cover all known insns (to allow it to be
used by more callers) without at the same time completing the actual
emulation logic.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Tue, 16 Aug 2016 13:35:13 +0000 (15:35 +0200)]
x86emul: drop RIP-relative special case for TEST
Moving ahead the "early operand adjustments" logic, the "test $imm,r/m"
special logic in the determination of the instruction boundary is no
longer necessary.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Mon, 15 Aug 2016 10:22:14 +0000 (12:22 +0200)]
x86emul: introduce SrcImm16
... and use it for RET, LRET, and ENTER processing to limit the amount
of "manual" insn bytes fetching. Note that for the RET and LRET paths
the change utilizes that SrcImplicit (aka SrcNone) table entries leave
src.val as zero.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Mon, 15 Aug 2016 08:41:48 +0000 (10:41 +0200)]
x86: force suitable alignment in sources rather than in linker script
Besides being more logical this also allows verifying correct recording
of alignments in .o files.
The cpu0_stack related ASSERT() in xen.lds.S is now of questionable
value (as it now verifies correct tool chain behavior), but I've left
it in nevertheless.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Mon, 15 Aug 2016 08:41:02 +0000 (10:41 +0200)]
build-id: fix minor quirks
The initial size check in xen_build_id_check() came too late (after the
first access to the structure), but was mostly redundant with checks
done in all callers; convert it to a properly placed ASSERT(). The
"mostly" part being addressed too: xen_build_init() was off by one.
And then there was a stray semicolon.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Jan Beulich [Fri, 12 Aug 2016 14:57:07 +0000 (16:57 +0200)]
x86emul: don't special case fetching unsigned 8-bit immediates
These can be made work using SrcImmByte, making sure the low 8 bits of
src.val get suitably zero extended upon consumption. SHLD and SHRD
require a little more adjustment: Their source operands get changed
away from SrcReg, handling the register access "manually" instead of
the insn byte fetching.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Fri, 12 Aug 2016 14:55:48 +0000 (16:55 +0200)]
x86emul: all push flavors are data moves
Make all paths leading to the "push" label have the Mov flag set, and
ASSERT() that to be the case. For the opcode FF group the adjustment is
benign for the paths not leading to "push", as they all set dst.type to
OP_NONE
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Fri, 12 Aug 2016 14:54:24 +0000 (16:54 +0200)]
x86emul: don't special case fetching the immediate of PUSH
These immediates follow the standard patterns in all modes, so they're
better fetched by the generic source operand handling code.
To facilitate testing, instead of adding yet another of these pretty
convoluted individual test cases, simply introduce another blowfish run
with -mno-accumulate-outgoing-args (the additional -Dstatic is to
keep the compiler from converting the calling convention to
"regparm(3)", which I did observe it does).
To make this introduction of a new blowfish pass (and potential further
ones later one) have less impact on the readability of the final code,
abstract all such "binary blob" executions via a table to iterate
through.
The resulting native code execution adjustment also uncovered a lack of
clobbers on the asm() in the 64-bit case, which is being fixed at once.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Razvan Cojocaru [Fri, 12 Aug 2016 14:51:36 +0000 (16:51 +0200)]
vm_event: synchronize vCPU state in vm_event_resume()
Vm_event_vcpu_pause() needs to use vcpu_pause_nosync() in order
for the current vCPU to not get stuck. A consequence of this is
that the custom vm_event response handlers will not always see
the real vCPU state in v->arch.user_regs. This patch makes sure
that the state is always synchronized in vm_event_resume, before
any handlers have been called. This problem especially affects
vm_event_set_registers().
Simply checking vm_event_pause_count to make sure the vCPU is
paused suffices since there's only one ring / consumer at a
time, and events are being processed one-by-one, so the
toolstack won't unpause the vCPU behind our backs.
Signed-off-by: Razvan Cojocaru <rcojocaru@bitdefender.com> Acked-by: Tamas K Lengyel <tamas@tklengyel.com>
Andrew Cooper [Thu, 11 Aug 2016 17:21:14 +0000 (17:21 +0000)]
x86/cpufreq: Avoid using processor_pminfo[cpu] when it is NULL
The undefined behaviour sanitiser shows that it really is NULL via the
pre_initcall path.
(XEN) ================================================================================
(XEN) UBSAN: Undefined behaviour in cpufreq.c:158:66
(XEN) member access within null pointer of type 'struct processor_pminfo'
(XEN) ----[ Xen-4.8-unstable x86_64 debug=y Not tainted ]----
<snip>
(XEN) [<ffff82d0801c4231>] cpufreq_add_cpu+0x161/0xdc0
(XEN) [<ffff82d0801c6610>] cpufreq.c#cpu_callback+0x20/0x30
(XEN) [<ffff82d0804eefad>] cpufreq.c#cpufreq_presmp_init+0x2d/0x50
(XEN) [<ffff82d0804c5942>] do_presmp_initcalls+0x22/0x30
(XEN) [<ffff82d08051852d>] __start_xen+0x378d/0x42f0
(XEN) [<ffff82d080100073>] __high_start+0x53/0x60
Fix two other occurances of the same buggy logic.
The processor_pminfo[] objects are only allocated as a result of
XENPF_set_processor_pminfo hypercalls, which means that this early cpu
callback will always hit the early NULL check, and is therefore pointless.
Signed-off-by: Andrew Cooper <andrew.cooper3@citrix.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
Jan Beulich [Thu, 11 Aug 2016 11:36:42 +0000 (13:36 +0200)]
x86/NUMA: cleanup
- drop the only left CONFIG_NUMA conditional (this is always true)
- drop struct node_data's node_id field (being always equal to the
node_data[] array index used)
- don't open code node_{start,end}_pfn() nor node_spanned_pages()
except when used as lvalues (those could be converted too, but this
seems a little awkward)
- no longer open code pfn_to_paddr() in an expression being modified
anyway
- make dump less verbose by logging actual vs intended node IDs only
when they don't match
Signed-off-by: Jan Beulich <jbeulich@suse.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Jan Beulich [Thu, 11 Aug 2016 11:35:50 +0000 (13:35 +0200)]
page-alloc/x86: don't restrict DMA heap to node 0
When node zero has no memory, the DMA bit width will end up getting set
to 9, which is obviously not helpful to hold back a reasonable amount
of low enough memory for Dom0 to use for DMA purposes. Find the lowest
node with memory below 4Gb instead.
Introduce arch_get_dma_bitsize() to keep this arch-specific logic out
of common code.
Also adjust the original calculation: I think the subtraction of 1
should have been part of the flsl() argument rather than getting
applied to its result. And while previously the division by 4 was valid
to be done on the flsl() result, this now also needs to be converted,
as is should only be applied to the spanned pages value.
Signed-off-by: Jan Beulich <jbeulich@suse.com> Acked-by: Julien Grall <julien.grall@arm.com> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>
Trammell Hudson [Thu, 11 Aug 2016 11:34:59 +0000 (13:34 +0200)]
allow reproducible builds of xen.gz
The mkelf32 executable was using an uninitialized stack buffer for
padding after the ehdr and phdr are written to the xen file, which
leads to non-deterministic bytes in the binary and prevented Xen
hypervisors from being reproducibly built.
Additionally, the file was then compressed with gzip -9 without the
-n | --no-name flag, which lead to the xen.gz file having
non-deterministric bytes (the timestamp) in the compressed file.
Signed-off-by: Trammell Hudson <trammell.hudson@twosigma.com> Reviewed-by: Jan Beulich <jbeulich@suse.com>
projects / xen.git / log