* [PATCH v4 00/16] Xen VMware tools support
@ 2014-09-11 18:36 Don Slutz
2014-09-11 18:36 ` [PATCH v4 01/16] xen: Add support for VMware cpuid leaves Don Slutz
` (15 more replies)
0 siblings, 16 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
Changes v3 to v4:
Ian Campbell:
Report on both viridian and vmware_hw set.
Added LIBXL_VGA_INTERFACE_TYPE_VMWARE (vga=vmware).
Andrew Cooper:
Add doc for hypervisor-cpuid.
Boris Ostrovsky:
Changing regs->error_code may not be a good idea.
Dropped this.
Jan Beulich & Boris Ostrovsky:
Only enable vmwxit for GP when vmware_port is set.
Done.
Changes v2 to v3:
Add optional unit test tools.
Re-worked split of changes.
Jan Beulich:
for #0:
I don't think you should be adding a new fine in hvm/ _and_ a new
subdirectory.
Moved all files to hvm/vmware that contain code.
for old #1 (now #1 & #2):
Is there really a point in enabling both Viridian and VMware extensions?
I still think so.
hvmloader change: This needs an explanation
Dropped as not need now.
Can you make vmware_hw similar to Viridian, returning success when
setting the value to what it already is.
Done.
You don't seem to be using sub_idx: ...
Dropped.
Extra changes...
Dropped.
for old #2 (now #3):
... these guards have the (theoretical at this point) risk of clashing
... the patch is obviously incomplete without this header...
Did not fix any of these issues. I will stick with this needs
to be a 2nd patch that changes the include files to better fit
in Xen coding. For now these files are in a sub directory
which is not part of the normal include search.
Moved the includeCheck.h file into this patch.
for old #3 (now #4, #5, #6, #7, #8, #9, #10, #11)
As I think was said on v1 already - this should be split into smaller
pieces ...
Done.
All this would very likely better go into a separate function placed in
vmport.c.
Moved most of the code into vmport.c or vmport_rpc.c.
In any event I'm rather uncomfortable about vmware_port getting
enabled unconditionally, ...
Added vmware_port (done in new patches #4, #5) as an xl.cfg
option.
You'll have to go through and fix coding style issues.
I think I have found all these, but since they do not stand out
for me, let me know of any left.
"MAKE_INSTR(IN," name is ambiguous.
Added all 4 opcodes for in and out that can access this port: INB_DX,
INL_DX, OUTB_DX, OUTL_DX.
A VMX-specific function shouldn't be named this way...
Added new common routine vmport_gp_check() that is called from
both vmx.c and svm.c which is where all the logic about checking
for IN ans OUT is done.
Also fixed naming and added static.
Ah, here we go (as to using HVM_DBG_LOG()): Isn't this _way_ too
fine grained?
I have reduced the number of bits used. Partialy by switching
some to xentrace (new patch #6 and #7).
Right, and zero is an indication that it wasn't found. Also I just
noticed there's a gdprintk() in that event, which for all other ...
Made the gdprintk() optional.
End of v3 changes.
This is a small part of the changes needed to allow running Linux
and windows (and others) guests that were built on VMware and run
run them unchanged on Xen.
This small part is the start of Xen support of VMware backdoor I/O
port which is how VMware tools (a standard addition installed on a
guest) communicates to the hypervisor.
I picked this subset to start with because it only has changes in
Xen.
Some of this code is already in QEMU and so KVM has some of this
already. QEMU supported backdoor commands include VMware mouse
support. A later patch set exists that links these changes, new
code and Xen changes to QEMU to provide VMware mouse support under
Xen. The important part is that VMware mouse is an absolute
position mouse and so network delays do not effect usage of the
virtual mouse.
For example from the guest:
[root@C63-min-tools ~]# vmtoolsd --cmd "info-get guestinfo.joejoel"
No value found
[root@C63-min-tools ~]# vmtoolsd --cmd "info-set guestinfo.joejoel short"
[root@C63-min-tools ~]# vmtoolsd --cmd "info-get guestinfo.joejoel"
short
[root@C63-min-tools ~]# vmtoolsd --cmd "info-set guestinfo.joejoel long222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222222200000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000joel"
[root@C63-min-tools ~]# vmtoolsd --cmd "info-get guestinfo.key1"
data1
[root@C63-min-tools ~]# vmtoolsd --cmd "info-get guestinfo.key2"
No value found
[root@C63-min-tools ~]# vmtoolsd --cmd "info-get guestinfo.key2"
data2
[root@C63-min-tools ~]#
Most of this code has been reverse engineered by looking at
source code for Linux and open VMware tools.
http://open-vm-tools.sourceforge.net
changes RFC to v2:
Jan Beulich:
Add xen/arch/x86/hvm/vmware.c for cpuid_vmware_leaves
Fewer patches
Andrew Cooper:
use the proper constant for apic_khz
Follow 839b966e3f587bbb1a0d954230fb3904330dccb6 style changes.
Changed HVM_PARAM_VMWARE_HW to write once (make is_vmware_domain()
more static).
Dropped vmport status stuff.
Added checks for xzalloc() having failed.
You should include backdoor_def.h ...
Every thing I tried did not work better. So I did not
change VMPORT_PORT and BDOOR_PORT being the same value.
I did not try and adjust VMware's include file backdoor_def.h
to working in other xen source files.
Switching to s_time_t is not valid. get_sec() is defined:
unsigned long get_sec(void);
and so my uses of it should be using unsigned long. However
since that is not a fixed width type, I used the uint64_t
data type which is almost the same, but does allow the 32 bit
build of libxc, libxl to do the correct thing.
Konrad Rzeszutek Wilk:
Please don't include the address. It should be, etc
about the Vmware provided include files.
I went with no changes to these files. Even if the files should
be changed to match xen coding style, etc I still feel that the
original ones should be added via a patch, and then adjusted in a
2nd patch.
Can you use XenBus?
I would say no. XenBus (and XenStore) is about domain to domain
communication. This is about VMware's hyper-call and providing
access to VMware's guest info very low speed access.
Olaf Hering:
Dropped changing of bios-strings. Still needs some documentation
about this may be needed to do in a tool stack or set of commands.
Boris Ostrovsky:
Use svm_nextrip_insn_length()
Looks like __get_instruction_length() does this, so switched to
__get_instruction_length().
RFC:
See
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009458
for info on detecting VMware.
Linux does not follow this exactly. It checks for CPUID 1st. If
that fails, it checks for SMBIOS containing "VMware" (not VMware- or
VMW).
So this patch set provides:
SMBIOS -- Add string VMware-
CPUID -- Add VMware's CPUID (Note: currently HyperV (viridian support) breaks this check.)
Add the magic VMware port
Allow VMware tools poweroff and reboot
Enable access to VMware's guest info
Provide the VMware tools build number
Don Slutz (16):
xen: Add support for VMware cpuid leaves
tools: Add vmware_hw support
vmware: Add VMware provided include files.
xen: Add is_vmware_port_enabled
tools: Add vmware_port support
xen: Convert vmware_port to xentrace usage
tools: Convert vmware_port to xentrace usage
xen: Add limited support of VMware's hyper-call rpc
tools: Add limited support of VMware's hyper-call rpc
Add VMware tool's triggers
Add live migration of VMware's hyper-call RPC
Add dump of HVM_SAVE_CODE(VMPORT) to xen-hvmctx.
Add xen-hvm-param
Add xen-vmware-guestinfo
Add xen-list-vmware-guestinfo
Add xen-hvm-send-trigger
.gitignore | 4 +
docs/man/xl.cfg.pod.5 | 28 +-
docs/misc/hypervisor-cpuid.markdown | 29 +
tools/libxc/xc_domain.c | 115 ++
tools/libxc/xc_domain_restore.c | 14 +
tools/libxc/xc_domain_save.c | 11 +
tools/libxc/xenctrl.h | 24 +
tools/libxc/xg_save_restore.h | 2 +
tools/libxl/libxl.h | 15 +
tools/libxl/libxl_create.c | 6 +-
tools/libxl/libxl_dm.c | 10 +-
tools/libxl/libxl_dom.c | 2 +
tools/libxl/libxl_types.idl | 3 +
tools/libxl/xl_cmdimpl.c | 12 +-
tools/misc/Makefile | 16 +-
tools/misc/xen-hvm-param.c | 164 +++
tools/misc/xen-hvm-send-trigger.c | 103 ++
tools/misc/xen-hvmctx.c | 229 ++++
tools/misc/xen-list-vmware-guestinfo.c | 88 ++
tools/misc/xen-vmware-guestinfo.c | 97 ++
tools/xentrace/formats | 13 +
xen/arch/x86/domain.c | 2 +
xen/arch/x86/domctl.c | 34 +
xen/arch/x86/hvm/Makefile | 3 +-
xen/arch/x86/hvm/hvm.c | 79 ++
xen/arch/x86/hvm/svm/emulate.c | 26 +-
xen/arch/x86/hvm/svm/svm.c | 53 +-
xen/arch/x86/hvm/svm/vmcb.c | 2 +
xen/arch/x86/hvm/vmware/Makefile | 3 +
xen/arch/x86/hvm/vmware/backdoor_def.h | 167 +++
xen/arch/x86/hvm/vmware/cpuid.c | 69 ++
xen/arch/x86/hvm/vmware/guest_msg_def.h | 87 ++
xen/arch/x86/hvm/vmware/includeCheck.h | 17 +
xen/arch/x86/hvm/vmware/vmport.c | 324 ++++++
xen/arch/x86/hvm/vmware/vmport_rpc.c | 1581 ++++++++++++++++++++++++++
xen/arch/x86/hvm/vmx/vmcs.c | 1 +
xen/arch/x86/hvm/vmx/vmx.c | 72 ++
xen/arch/x86/hvm/vmx/vvmx.c | 14 +
xen/arch/x86/traps.c | 8 +-
xen/common/domctl.c | 3 +
xen/include/asm-x86/hvm/domain.h | 7 +
xen/include/asm-x86/hvm/hvm.h | 3 +
xen/include/asm-x86/hvm/io.h | 2 +-
xen/include/asm-x86/hvm/svm/emulate.h | 13 +-
xen/include/asm-x86/hvm/trace.h | 45 +
xen/include/asm-x86/hvm/vmport.h | 72 ++
xen/include/asm-x86/hvm/vmware.h | 31 +
xen/include/public/arch-x86/hvm/save.h | 39 +-
xen/include/public/arch-x86/hvm/vmporttype.h | 118 ++
xen/include/public/domctl.h | 6 +
xen/include/public/hvm/hvm_op.h | 18 +
xen/include/public/hvm/params.h | 8 +-
xen/include/public/trace.h | 12 +
xen/include/xen/sched.h | 3 +
54 files changed, 3884 insertions(+), 23 deletions(-)
create mode 100644 docs/misc/hypervisor-cpuid.markdown
create mode 100644 tools/misc/xen-hvm-param.c
create mode 100644 tools/misc/xen-hvm-send-trigger.c
create mode 100644 tools/misc/xen-list-vmware-guestinfo.c
create mode 100644 tools/misc/xen-vmware-guestinfo.c
create mode 100644 xen/arch/x86/hvm/vmware/Makefile
create mode 100644 xen/arch/x86/hvm/vmware/backdoor_def.h
create mode 100644 xen/arch/x86/hvm/vmware/cpuid.c
create mode 100644 xen/arch/x86/hvm/vmware/guest_msg_def.h
create mode 100644 xen/arch/x86/hvm/vmware/includeCheck.h
create mode 100644 xen/arch/x86/hvm/vmware/vmport.c
create mode 100644 xen/arch/x86/hvm/vmware/vmport_rpc.c
create mode 100644 xen/include/asm-x86/hvm/vmport.h
create mode 100644 xen/include/asm-x86/hvm/vmware.h
create mode 100644 xen/include/public/arch-x86/hvm/vmporttype.h
--
1.8.4
^ permalink raw reply [flat|nested] 51+ messages in thread
* [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 19:49 ` Andrew Cooper
2014-09-12 12:37 ` Boris Ostrovsky
2014-09-11 18:36 ` [PATCH v4 02/16] tools: Add vmware_hw support Don Slutz
` (14 subsequent siblings)
15 siblings, 2 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
This is done by adding HVM_PARAM_VMWARE_HW. It is set to the VMware
virtual hardware version.
Currently 0, 3-4, 6-11 are good values. However the
code only checks for == 0 or != 0.
If non-zero then
Return VMware's cpuid leaves.
The support of hypervisor cpuid leaves has not been agreed to.
MicroSoft Hyper-V (AKA viridian) currently must be at 0x40000000.
VMware currently must be at 0x40000000.
KVM currently must be at 0x40000000 (from Seabios).
Seabios will find xen at 0x40000000, 0x40000100, 0x40000200 ..
0x40010000.
http://download.microsoft.com/download/F/B/0/FB0D01A3-8E3A-4F5F-AA59-08C8026D3B8A/requirements-for-implementing-microsoft-hypervisor-interface.docx
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009458
http://lwn.net/Articles/301888/
Attempted to get this cleaned up.
So based on this, I picked the order:
Xen at 0x40000000 or
Viridian or VMware at 0x40000000 and Xen at 0x40000100
If both Viridian and VMware selected, report an error.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
xen/arch/x86/hvm/Makefile | 3 +-
xen/arch/x86/hvm/hvm.c | 32 +++++++++++++++++++
xen/arch/x86/hvm/vmware/Makefile | 1 +
xen/arch/x86/hvm/vmware/cpuid.c | 69 ++++++++++++++++++++++++++++++++++++++++
xen/arch/x86/traps.c | 8 +++--
xen/include/asm-x86/hvm/hvm.h | 3 ++
xen/include/asm-x86/hvm/vmware.h | 31 ++++++++++++++++++
xen/include/public/hvm/params.h | 5 ++-
8 files changed, 148 insertions(+), 4 deletions(-)
create mode 100644 xen/arch/x86/hvm/vmware/Makefile
create mode 100644 xen/arch/x86/hvm/vmware/cpuid.c
create mode 100644 xen/include/asm-x86/hvm/vmware.h
diff --git a/xen/arch/x86/hvm/Makefile b/xen/arch/x86/hvm/Makefile
index eea5555..77598a6 100644
--- a/xen/arch/x86/hvm/Makefile
+++ b/xen/arch/x86/hvm/Makefile
@@ -1,5 +1,6 @@
subdir-y += svm
subdir-y += vmx
+subdir-y += vmware
obj-y += asid.o
obj-y += emulate.o
@@ -22,4 +23,4 @@ obj-y += vlapic.o
obj-y += vmsi.o
obj-y += vpic.o
obj-y += vpt.o
-obj-y += vpmu.o
\ No newline at end of file
+obj-y += vpmu.o
diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 8d905d3..03a1a19 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -57,6 +57,7 @@
#include <asm/hvm/cacheattr.h>
#include <asm/hvm/trace.h>
#include <asm/hvm/nestedhvm.h>
+#include <asm/hvm/vmware.h>
#include <asm/mtrr.h>
#include <asm/apic.h>
#include <public/sched.h>
@@ -4228,6 +4229,9 @@ void hvm_cpuid(unsigned int input, unsigned int *eax, unsigned int *ebx,
if ( cpuid_viridian_leaves(input, eax, ebx, ecx, edx) )
return;
+ if ( cpuid_vmware_leaves(input, eax, ebx, ecx, edx) )
+ return;
+
if ( cpuid_hypervisor_leaves(input, count, eax, ebx, ecx, edx) )
return;
@@ -5555,6 +5559,11 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
rc = -EINVAL;
break;
case HVM_PARAM_VIRIDIAN:
+ if ( d->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW] )
+ {
+ rc = -EXDEV;
+ break;
+ }
if ( a.value > 1 )
rc = -EINVAL;
break;
@@ -5692,6 +5701,29 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
break;
}
+ case HVM_PARAM_VMWARE_HW:
+ /*
+ * This should only ever be set non-zero one time by
+ * the tools and is read only by the guest.
+ */
+ if ( d == current->domain )
+ {
+ rc = -EPERM;
+ break;
+ }
+ if ( d->arch.hvm_domain.params[HVM_PARAM_VIRIDIAN] )
+ {
+ rc = -EXDEV;
+ break;
+ }
+ if ( d->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW] &&
+ d->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW] !=
+ a.value )
+ {
+ rc = -EEXIST;
+ break;
+ }
+ break;
}
if ( rc == 0 )
diff --git a/xen/arch/x86/hvm/vmware/Makefile b/xen/arch/x86/hvm/vmware/Makefile
new file mode 100644
index 0000000..3fb2e0b
--- /dev/null
+++ b/xen/arch/x86/hvm/vmware/Makefile
@@ -0,0 +1 @@
+obj-y += cpuid.o
diff --git a/xen/arch/x86/hvm/vmware/cpuid.c b/xen/arch/x86/hvm/vmware/cpuid.c
new file mode 100644
index 0000000..730ab8f
--- /dev/null
+++ b/xen/arch/x86/hvm/vmware/cpuid.c
@@ -0,0 +1,69 @@
+/*
+ * arch/x86/hvm/vmware/cpuid.c
+ *
+ * Copyright (C) 2012 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#include <xen/sched.h>
+
+#include <asm/hvm/hvm.h>
+#include <asm/hvm/vmware.h>
+
+int cpuid_vmware_leaves(uint32_t idx, uint32_t *eax, uint32_t *ebx,
+ uint32_t *ecx, uint32_t *edx)
+{
+ struct domain *d = current->domain;
+
+ if ( !is_vmware_domain(d) )
+ return 0;
+
+ switch ( idx - 0x40000000 )
+ {
+ case 0x0:
+ *eax = 0x40000010; /* Largest leaf */
+ *ebx = 0x61774d56; /* "VMwa" */
+ *ecx = 0x4d566572; /* "reVM" */
+ *edx = 0x65726177; /* "ware" */
+ break;
+
+ case 0x1 ... 0xf:
+ *eax = 0; /* Reserved */
+ *ebx = 0; /* Reserved */
+ *ecx = 0; /* Reserved */
+ *edx = 0; /* Reserved */
+ break;
+
+ case 0x10:
+ /* (Virtual) TSC frequency in kHz. */
+ *eax = d->arch.tsc_khz;
+ /* (Virtual) Bus (local apic timer) frequency in kHz. */
+ *ebx = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
+ *ecx = 0; /* Reserved */
+ *edx = 0; /* Reserved */
+ break;
+
+ default:
+ return 0;
+ }
+
+ return 1;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
index 10fc2ca..f353f42 100644
--- a/xen/arch/x86/traps.c
+++ b/xen/arch/x86/traps.c
@@ -685,8 +685,12 @@ int cpuid_hypervisor_leaves( uint32_t idx, uint32_t sub_idx,
uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx)
{
struct domain *d = current->domain;
- /* Optionally shift out of the way of Viridian architectural leaves. */
- uint32_t base = is_viridian_domain(d) ? 0x40000100 : 0x40000000;
+ /*
+ * Optionally shift out of the way of Viridian or VMware
+ * architectural leaves.
+ */
+ uint32_t base = is_viridian_domain(d) | is_vmware_domain(d) ?
+ 0x40000100 : 0x40000000;
uint32_t limit, dummy;
idx -= base;
diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
index 1123857..546210a 100644
--- a/xen/include/asm-x86/hvm/hvm.h
+++ b/xen/include/asm-x86/hvm/hvm.h
@@ -347,6 +347,9 @@ static inline unsigned long hvm_get_shadow_gs_base(struct vcpu *v)
#define is_viridian_domain(_d) \
(is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VIRIDIAN]))
+#define is_vmware_domain(_d) \
+ (is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW]))
+
void hvm_hypervisor_cpuid_leaf(uint32_t sub_idx,
uint32_t *eax, uint32_t *ebx,
uint32_t *ecx, uint32_t *edx);
diff --git a/xen/include/asm-x86/hvm/vmware.h b/xen/include/asm-x86/hvm/vmware.h
new file mode 100644
index 0000000..f254106
--- /dev/null
+++ b/xen/include/asm-x86/hvm/vmware.h
@@ -0,0 +1,31 @@
+/*
+ * asm-x86/hvm/vmware.h
+ *
+ * Copyright (C) 2012 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef ASM_X86_HVM_VMWARE_H__
+#define ASM_X86_HVM_VMWARE_H__
+
+int cpuid_vmware_leaves(uint32_t idx, uint32_t *eax, uint32_t *ebx,
+ uint32_t *ecx, uint32_t *edx);
+
+#endif /* ASM_X86_HVM_VMWARE_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/public/hvm/params.h b/xen/include/public/hvm/params.h
index 614ff5f..dee6d68 100644
--- a/xen/include/public/hvm/params.h
+++ b/xen/include/public/hvm/params.h
@@ -151,6 +151,9 @@
/* Location of the VM Generation ID in guest physical address space. */
#define HVM_PARAM_VM_GENERATION_ID_ADDR 34
-#define HVM_NR_PARAMS 35
+/* Params for VMware */
+#define HVM_PARAM_VMWARE_HW 35
+
+#define HVM_NR_PARAMS 36
#endif /* __XEN_PUBLIC_HVM_PARAMS_H__ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 02/16] tools: Add vmware_hw support
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
2014-09-11 18:36 ` [PATCH v4 01/16] xen: Add support for VMware cpuid leaves Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 21:09 ` Andrew Cooper
2014-09-11 18:36 ` [PATCH v4 03/16] vmware: Add VMware provided include files Don Slutz
` (13 subsequent siblings)
15 siblings, 1 reply; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
This is used to set HVM_PARAM_VMWARE_HW. It is set to the VMware
virtual hardware version.
Currently 0, 3-4, 6-11 are good values. However the code only
checks for == 0 or != 0.
If non-zero then
default VGA to VMware's VGA.
Also now allows vga=vmware
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
docs/man/xl.cfg.pod.5 | 21 +++++++++++++++++++--
docs/misc/hypervisor-cpuid.markdown | 29 +++++++++++++++++++++++++++++
tools/libxc/xc_domain_restore.c | 14 ++++++++++++++
tools/libxc/xc_domain_save.c | 11 +++++++++++
tools/libxc/xg_save_restore.h | 2 ++
tools/libxl/libxl.h | 10 ++++++++++
tools/libxl/libxl_create.c | 4 +++-
tools/libxl/libxl_dm.c | 10 +++++++++-
tools/libxl/libxl_dom.c | 2 ++
tools/libxl/libxl_types.idl | 2 ++
tools/libxl/xl_cmdimpl.c | 11 ++++++++++-
11 files changed, 111 insertions(+), 5 deletions(-)
create mode 100644 docs/misc/hypervisor-cpuid.markdown
diff --git a/docs/man/xl.cfg.pod.5 b/docs/man/xl.cfg.pod.5
index 517ae2f..367b401 100644
--- a/docs/man/xl.cfg.pod.5
+++ b/docs/man/xl.cfg.pod.5
@@ -1147,6 +1147,23 @@ some other Operating Systems and in some circumstance can prevent
Xen's own paravirtualisation interfaces for HVM guests from being
used.
+=item B<vmware_hw=NUMBER>
+
+Turns on or off the exposure of VMware cpuid. The number is the
+VMware's hardware version number, where 0 is off. If not zero it
+changes the default VGA to VMware's VGA.
+
+The hardware version number (vmware_hw) come from VMware config files.
+
+=over 4
+
+In a .vmx it is virtualHW.version
+
+In a .ovf it is part of the value of vssd:VirtualSystemType.
+For vssd:VirtualSystemType == vmx-07, vmware_hw = 7.
+
+=back
+
=back
=head3 Emulated VGA Graphics Device
@@ -1185,8 +1202,8 @@ This option is deprecated, use vga="stdvga" instead.
=item B<vga="STRING">
-Selects the emulated video card (none|stdvga|cirrus).
-The default is cirrus.
+Selects the emulated video card (none|stdvga|cirrus|vmware).
+The default is cirrus (or vmware if B<vmware_hw> is not zero).
=item B<vnc=BOOLEAN>
diff --git a/docs/misc/hypervisor-cpuid.markdown b/docs/misc/hypervisor-cpuid.markdown
new file mode 100644
index 0000000..4199d57
--- /dev/null
+++ b/docs/misc/hypervisor-cpuid.markdown
@@ -0,0 +1,29 @@
+Hypervisor Cpuid
+================
+
+The support of hypervisor cpuid leaves has not been agreed to.
+Other then the range 0x40000000 to 0x400000ff can be used by
+hypervisors.
+
+MicroSoft Hyper-V (AKA viridian) currently must be at 0x40000000.
+
+VMware currently must be at 0x40000000.
+
+KVM currently must be at 0x40000000 (from Seabios).
+
+Seabios will find Xen at 0x40000000, 0x40000100, 0x40000200 ..
+0x40010000.
+
+http://download.microsoft.com/download/F/B/0/FB0D01A3-8E3A-4F5F-AA59-08C8026D3B8A/requirements-for-implementing-microsoft-hypervisor-interface.docx
+
+http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009458
+
+http://lwn.net/Articles/301888/
+ Attempted to get this cleaned up.
+
+So if Viridian or VMware_hw is selected, return their format for the
+range 0x40000000 to 0x400000ff. And return Xen format for the range
+0x40000100 to 0x400001ff.
+
+Otherwise return Xen format for the range 0x40000000 to 0x400000ff.
+
diff --git a/tools/libxc/xc_domain_restore.c b/tools/libxc/xc_domain_restore.c
index b9a56d5..bc5cd57 100644
--- a/tools/libxc/xc_domain_restore.c
+++ b/tools/libxc/xc_domain_restore.c
@@ -743,6 +743,7 @@ typedef struct {
uint64_t vm_generationid_addr;
uint64_t ioreq_server_pfn;
uint64_t nr_ioreq_server_pages;
+ uint64_t vmware_hw;
struct toolstack_data_t tdata;
} pagebuf_t;
@@ -927,6 +928,16 @@ static int pagebuf_get_one(xc_interface *xch, struct restore_ctx *ctx,
}
return pagebuf_get_one(xch, ctx, buf, fd, dom);
+ case XC_SAVE_ID_HVM_VMWARE_HW:
+ /* Skip padding 4 bytes then read the vmware hw version. */
+ if ( RDEXACT(fd, &buf->vmware_hw, sizeof(uint32_t)) ||
+ RDEXACT(fd, &buf->vmware_hw, sizeof(uint64_t)) )
+ {
+ PERROR("error read the vmware_hw value");
+ return -1;
+ }
+ return pagebuf_get_one(xch, ctx, buf, fd, dom);
+
case XC_SAVE_ID_TOOLSTACK:
{
if ( RDEXACT(fd, &buf->tdata.len, sizeof(buf->tdata.len)) )
@@ -1760,6 +1771,9 @@ int xc_domain_restore(xc_interface *xch, int io_fd, uint32_t dom,
}
}
+ if (pagebuf.vmware_hw != 0)
+ xc_set_hvm_param(xch, dom, HVM_PARAM_VMWARE_HW, pagebuf.vmware_hw);
+
if (pagebuf.acpi_ioport_location == 1) {
DBGPRINTF("Use new firmware ioport from the checkpoint\n");
xc_hvm_param_set(xch, dom, HVM_PARAM_ACPI_IOPORTS_LOCATION, 1);
diff --git a/tools/libxc/xc_domain_save.c b/tools/libxc/xc_domain_save.c
index 254fdb3..76dc307 100644
--- a/tools/libxc/xc_domain_save.c
+++ b/tools/libxc/xc_domain_save.c
@@ -1750,6 +1750,17 @@ int xc_domain_save(xc_interface *xch, int io_fd, uint32_t dom, uint32_t max_iter
PERROR("Error when writing the ioreq server gmfn count");
goto out;
}
+
+ chunk.id = XC_SAVE_ID_HVM_VMWARE_HW;
+ chunk.data = 0;
+ xc_hvm_param_get(xch, dom, HVM_PARAM_VMWARE_HW, &chunk.data);
+
+ if ( (chunk.data != 0) &&
+ wrexact(io_fd, &chunk, sizeof(chunk)) )
+ {
+ PERROR("Error when writing the vmware_hw value");
+ goto out;
+ }
}
if ( callbacks != NULL && callbacks->toolstack_save != NULL )
diff --git a/tools/libxc/xg_save_restore.h b/tools/libxc/xg_save_restore.h
index bdd9009..d185ba9 100644
--- a/tools/libxc/xg_save_restore.h
+++ b/tools/libxc/xg_save_restore.h
@@ -262,6 +262,8 @@
/* These are a pair; it is an error for one to exist without the other */
#define XC_SAVE_ID_HVM_IOREQ_SERVER_PFN -19
#define XC_SAVE_ID_HVM_NR_IOREQ_SERVER_PAGES -20
+/* VMware data */
+#define XC_SAVE_ID_HVM_VMWARE_HW -21
/*
** We process save/restore/migrate in batches of pages; the below
diff --git a/tools/libxl/libxl.h b/tools/libxl/libxl.h
index dab3a67..067af5b 100644
--- a/tools/libxl/libxl.h
+++ b/tools/libxl/libxl.h
@@ -138,6 +138,16 @@
#define LIBXL_HAVE_BUILDINFO_IOMEM_START_GFN 1
/*
+ * The libxl_vga_interface_type has the type for vmware.
+ */
+#define LIBXL_HAVE_LIBXL_VGA_INTERFACE_TYPE_VMWARE 1
+
+/*
+ * libxl_domain_build_info has the u.hvm.vmware_hw field.
+ */
+#define LIBXL_HAVE_BUILDINFO_HVM_VMWARE_HW 1
+
+/*
* libxl ABI compatibility
*
* The only guarantee which libxl makes regarding ABI compatibility
diff --git a/tools/libxl/libxl_create.c b/tools/libxl/libxl_create.c
index 4533352..d9b60af 100644
--- a/tools/libxl/libxl_create.c
+++ b/tools/libxl/libxl_create.c
@@ -428,13 +428,15 @@ int libxl__domain_build(libxl__gc *gc,
vments[4] = "start_time";
vments[5] = libxl__sprintf(gc, "%lu.%02d", start_time.tv_sec,(int)start_time.tv_usec/10000);
- localents = libxl__calloc(gc, 7, sizeof(char *));
+ localents = libxl__calloc(gc, 9, sizeof(char *));
localents[0] = "platform/acpi";
localents[1] = libxl_defbool_val(info->u.hvm.acpi) ? "1" : "0";
localents[2] = "platform/acpi_s3";
localents[3] = libxl_defbool_val(info->u.hvm.acpi_s3) ? "1" : "0";
localents[4] = "platform/acpi_s4";
localents[5] = libxl_defbool_val(info->u.hvm.acpi_s4) ? "1" : "0";
+ localents[6] = "platform/vmware_hw";
+ localents[7] = libxl__sprintf(gc, "%"PRId64, info->u.hvm.vmware_hw);
break;
case LIBXL_DOMAIN_TYPE_PV:
diff --git a/tools/libxl/libxl_dm.c b/tools/libxl/libxl_dm.c
index 103cbca..d27b364 100644
--- a/tools/libxl/libxl_dm.c
+++ b/tools/libxl/libxl_dm.c
@@ -243,6 +243,9 @@ static char ** libxl__build_device_model_args_old(libxl__gc *gc,
case LIBXL_VGA_INTERFACE_TYPE_NONE:
flexarray_append_pair(dm_args, "-vga", "none");
break;
+ case LIBXL_VGA_INTERFACE_TYPE_VMWARE:
+ flexarray_append_pair(dm_args, "-vga", "vmware");
+ break;
}
if (b_info->u.hvm.boot) {
@@ -555,7 +558,12 @@ static char ** libxl__build_device_model_args_new(libxl__gc *gc,
break;
case LIBXL_VGA_INTERFACE_TYPE_NONE:
break;
- }
+ case LIBXL_VGA_INTERFACE_TYPE_VMWARE:
+ flexarray_append_pair(dm_args, "-device",
+ GCSPRINTF("vmware-svga,vgamem_mb=%d",
+ libxl__sizekb_to_mb(b_info->video_memkb)));
+ break;
+ }
if (b_info->u.hvm.boot) {
flexarray_vappend(dm_args, "-boot",
diff --git a/tools/libxl/libxl_dom.c b/tools/libxl/libxl_dom.c
index 0dfdb08..ee805ca 100644
--- a/tools/libxl/libxl_dom.c
+++ b/tools/libxl/libxl_dom.c
@@ -219,6 +219,8 @@ static void hvm_set_conf_params(xc_interface *handle, uint32_t domid,
libxl_defbool_val(info->u.hvm.viridian));
xc_hvm_param_set(handle, domid, HVM_PARAM_HPET_ENABLED,
libxl_defbool_val(info->u.hvm.hpet));
+ xc_set_hvm_param(handle, domid, HVM_PARAM_VMWARE_HW,
+ info->u.hvm.vmware_hw);
#endif
xc_hvm_param_set(handle, domid, HVM_PARAM_TIMER_MODE, timer_mode(info));
xc_hvm_param_set(handle, domid, HVM_PARAM_VPT_ALIGN,
diff --git a/tools/libxl/libxl_types.idl b/tools/libxl/libxl_types.idl
index f1fcbc3..907572c 100644
--- a/tools/libxl/libxl_types.idl
+++ b/tools/libxl/libxl_types.idl
@@ -172,6 +172,7 @@ libxl_vga_interface_type = Enumeration("vga_interface_type", [
(1, "CIRRUS"),
(2, "STD"),
(3, "NONE"),
+ (4, "VMWARE"),
], init_val = "LIBXL_VGA_INTERFACE_TYPE_CIRRUS")
libxl_vendor_device = Enumeration("vendor_device", [
@@ -378,6 +379,7 @@ libxl_domain_build_info = Struct("domain_build_info",[
("timeoffset", string),
("hpet", libxl_defbool),
("vpt_align", libxl_defbool),
+ ("vmware_hw", UInt(64, init_val = 0)),
("timer_mode", libxl_timer_mode),
("nested_hvm", libxl_defbool),
("smbios_firmware", string),
diff --git a/tools/libxl/xl_cmdimpl.c b/tools/libxl/xl_cmdimpl.c
index 8a38077..db86382 100644
--- a/tools/libxl/xl_cmdimpl.c
+++ b/tools/libxl/xl_cmdimpl.c
@@ -1036,6 +1036,8 @@ static void parse_config_data(const char *config_source,
xlu_cfg_get_defbool(config, "hpet", &b_info->u.hvm.hpet, 0);
xlu_cfg_get_defbool(config, "vpt_align", &b_info->u.hvm.vpt_align, 0);
+ if (!xlu_cfg_get_long(config, "vmware_hw", &l, 1))
+ b_info->u.hvm.vmware_hw = l;
if (!xlu_cfg_get_long(config, "timer_mode", &l, 1)) {
const char *s = libxl_timer_mode_to_string(l);
fprintf(stderr, "WARNING: specifying \"timer_mode\" as an integer is deprecated. "
@@ -1674,13 +1676,20 @@ skip_vfb:
b_info->u.hvm.vga.kind = LIBXL_VGA_INTERFACE_TYPE_CIRRUS;
} else if (!strcmp(buf, "none")) {
b_info->u.hvm.vga.kind = LIBXL_VGA_INTERFACE_TYPE_NONE;
+ } else if (!strcmp(buf, "vmware")) {
+ b_info->u.hvm.vga.kind = LIBXL_VGA_INTERFACE_TYPE_VMWARE;
} else {
fprintf(stderr, "Unknown vga \"%s\" specified\n", buf);
exit(1);
}
} else if (!xlu_cfg_get_long(config, "stdvga", &l, 0))
b_info->u.hvm.vga.kind = l ? LIBXL_VGA_INTERFACE_TYPE_STD :
- LIBXL_VGA_INTERFACE_TYPE_CIRRUS;
+ b_info->u.hvm.vmware_hw ? LIBXL_VGA_INTERFACE_TYPE_VMWARE :
+ LIBXL_VGA_INTERFACE_TYPE_CIRRUS;
+ else
+ b_info->u.hvm.vga.kind =
+ b_info->u.hvm.vmware_hw ? LIBXL_VGA_INTERFACE_TYPE_VMWARE :
+ LIBXL_VGA_INTERFACE_TYPE_CIRRUS;
xlu_cfg_replace_string (config, "keymap", &b_info->u.hvm.keymap, 0);
xlu_cfg_get_defbool (config, "spice", &b_info->u.hvm.spice.enable, 0);
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 03/16] vmware: Add VMware provided include files.
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
2014-09-11 18:36 ` [PATCH v4 01/16] xen: Add support for VMware cpuid leaves Don Slutz
2014-09-11 18:36 ` [PATCH v4 02/16] tools: Add vmware_hw support Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 04/16] xen: Add is_vmware_port_enabled Don Slutz
` (12 subsequent siblings)
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
These 2 files: backdoor_def.h and guest_msg_def.h come from:
http://packages.vmware.com/tools/esx/3.5latest/rhel4/SRPMS/index.html
open-vm-tools-kmod-7.4.8-396269.423167.src.rpm
open-vm-tools-kmod-7.4.8.tar.gz
vmhgfs/backdoor_def.h
vmhgfs/guest_msg_def.h
and are unchanged.
Added the badly named include file includeCheck.h also. It only
has comments and is provided so that backdoor_def.h and
guest_msg_def.h can be used without change.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
xen/arch/x86/hvm/vmware/backdoor_def.h | 167 ++++++++++++++++++++++++++++++++
xen/arch/x86/hvm/vmware/guest_msg_def.h | 87 +++++++++++++++++
xen/arch/x86/hvm/vmware/includeCheck.h | 17 ++++
3 files changed, 271 insertions(+)
create mode 100644 xen/arch/x86/hvm/vmware/backdoor_def.h
create mode 100644 xen/arch/x86/hvm/vmware/guest_msg_def.h
create mode 100644 xen/arch/x86/hvm/vmware/includeCheck.h
diff --git a/xen/arch/x86/hvm/vmware/backdoor_def.h b/xen/arch/x86/hvm/vmware/backdoor_def.h
new file mode 100644
index 0000000..e76795f
--- /dev/null
+++ b/xen/arch/x86/hvm/vmware/backdoor_def.h
@@ -0,0 +1,167 @@
+/* **********************************************************
+ * Copyright 1998 VMware, Inc. All rights reserved.
+ * **********************************************************
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation version 2 and no later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/*
+ * backdoor_def.h --
+ *
+ * This contains backdoor defines that can be included from
+ * an assembly language file.
+ */
+
+
+
+#ifndef _BACKDOOR_DEF_H_
+#define _BACKDOOR_DEF_H_
+
+#define INCLUDE_ALLOW_MODULE
+#define INCLUDE_ALLOW_USERLEVEL
+#define INCLUDE_ALLOW_VMMEXT
+#define INCLUDE_ALLOW_VMCORE
+#define INCLUDE_ALLOW_VMKERNEL
+#include "includeCheck.h"
+
+/*
+ * If you want to add a new low-level backdoor call for a guest userland
+ * application, please consider using the GuestRpc mechanism instead. --hpreg
+ */
+
+#define BDOOR_MAGIC 0x564D5868
+
+/* Low-bandwidth backdoor port. --hpreg */
+
+#define BDOOR_PORT 0x5658
+
+#define BDOOR_CMD_GETMHZ 1
+/*
+ * BDOOR_CMD_APMFUNCTION is used by:
+ *
+ * o The FrobOS code, which instead should either program the virtual chipset
+ * (like the new BIOS code does, matthias offered to implement that), or not
+ * use any VM-specific code (which requires that we correctly implement
+ * "power off on CLI HLT" for SMP VMs, boris offered to implement that)
+ *
+ * o The old BIOS code, which will soon be jettisoned
+ *
+ * --hpreg
+ */
+#define BDOOR_CMD_APMFUNCTION 2
+#define BDOOR_CMD_GETDISKGEO 3
+#define BDOOR_CMD_GETPTRLOCATION 4
+#define BDOOR_CMD_SETPTRLOCATION 5
+#define BDOOR_CMD_GETSELLENGTH 6
+#define BDOOR_CMD_GETNEXTPIECE 7
+#define BDOOR_CMD_SETSELLENGTH 8
+#define BDOOR_CMD_SETNEXTPIECE 9
+#define BDOOR_CMD_GETVERSION 10
+#define BDOOR_CMD_GETDEVICELISTELEMENT 11
+#define BDOOR_CMD_TOGGLEDEVICE 12
+#define BDOOR_CMD_GETGUIOPTIONS 13
+#define BDOOR_CMD_SETGUIOPTIONS 14
+#define BDOOR_CMD_GETSCREENSIZE 15
+#define BDOOR_CMD_MONITOR_CONTROL 16
+#define BDOOR_CMD_GETHWVERSION 17
+#define BDOOR_CMD_OSNOTFOUND 18
+#define BDOOR_CMD_GETUUID 19
+#define BDOOR_CMD_GETMEMSIZE 20
+#define BDOOR_CMD_HOSTCOPY 21 /* Devel only */
+/* BDOOR_CMD_GETOS2INTCURSOR, 22, is very old and defunct. Reuse. */
+#define BDOOR_CMD_GETTIME 23 /* Deprecated. Use GETTIMEFULL. */
+#define BDOOR_CMD_STOPCATCHUP 24
+#define BDOOR_CMD_PUTCHR 25 /* Devel only */
+#define BDOOR_CMD_ENABLE_MSG 26 /* Devel only */
+#define BDOOR_CMD_GOTO_TCL 27 /* Devel only */
+#define BDOOR_CMD_INITPCIOPROM 28
+#define BDOOR_CMD_INT13 29
+#define BDOOR_CMD_MESSAGE 30
+#define BDOOR_CMD_RSVD0 31
+#define BDOOR_CMD_RSVD1 32
+#define BDOOR_CMD_RSVD2 33
+#define BDOOR_CMD_ISACPIDISABLED 34
+#define BDOOR_CMD_TOE 35 /* Not in use */
+/* BDOOR_CMD_INITLSIOPROM, 36, was merged with 28. Reuse. */
+#define BDOOR_CMD_PATCH_SMBIOS_STRUCTS 37
+#define BDOOR_CMD_MAPMEM 38 /* Devel only */
+#define BDOOR_CMD_ABSPOINTER_DATA 39
+#define BDOOR_CMD_ABSPOINTER_STATUS 40
+#define BDOOR_CMD_ABSPOINTER_COMMAND 41
+#define BDOOR_CMD_TIMER_SPONGE 42
+#define BDOOR_CMD_PATCH_ACPI_TABLES 43
+/* Catch-all to allow synchronous tests */
+#define BDOOR_CMD_DEVEL_FAKEHARDWARE 44 /* Debug only - needed in beta */
+#define BDOOR_CMD_GETHZ 45
+#define BDOOR_CMD_GETTIMEFULL 46
+#define BDOOR_CMD_STATELOGGER 47
+#define BDOOR_CMD_CHECKFORCEBIOSSETUP 48
+#define BDOOR_CMD_LAZYTIMEREMULATION 49
+#define BDOOR_CMD_BIOSBBS 50
+#define BDOOR_CMD_MAX 51
+
+/*
+ * IMPORTANT NOTE: When modifying the behavior of an existing backdoor command,
+ * you must adhere to the semantics expected by the oldest Tools who use that
+ * command. Specifically, do not alter the way in which the command modifies
+ * the registers. Otherwise backwards compatibility will suffer.
+ */
+
+/* High-bandwidth backdoor port. --hpreg */
+
+#define BDOORHB_PORT 0x5659
+
+#define BDOORHB_CMD_MESSAGE 0
+#define BDOORHB_CMD_MAX 1
+
+/*
+ * There is another backdoor which allows access to certain TSC-related
+ * values using otherwise illegal PMC indices when the pseudo_perfctr
+ * control flag is set.
+ */
+
+#define BDOOR_PMC_HW_TSC 0x10000
+#define BDOOR_PMC_REAL_NS 0x10001
+#define BDOOR_PMC_APPARENT_NS 0x10002
+
+#define IS_BDOOR_PMC(index) (((index) | 3) == 0x10003)
+#define BDOOR_CMD(ecx) ((ecx) & 0xffff)
+
+
+#ifdef VMM
+/*
+ *----------------------------------------------------------------------
+ *
+ * Backdoor_CmdRequiresFullyValidVCPU --
+ *
+ * A few backdoor commands require the full VCPU to be valid
+ * (including GDTR, IDTR, TR and LDTR). The rest get read/write
+ * access to GPRs and read access to Segment registers (selectors).
+ *
+ * Result:
+ * True iff VECX contains a command that require the full VCPU to
+ * be valid.
+ *
+ *----------------------------------------------------------------------
+ */
+static INLINE Bool
+Backdoor_CmdRequiresFullyValidVCPU(unsigned cmd)
+{
+ return cmd == BDOOR_CMD_RSVD0 ||
+ cmd == BDOOR_CMD_RSVD1 ||
+ cmd == BDOOR_CMD_RSVD2;
+}
+#endif
+
+#endif
diff --git a/xen/arch/x86/hvm/vmware/guest_msg_def.h b/xen/arch/x86/hvm/vmware/guest_msg_def.h
new file mode 100644
index 0000000..44ae0fa
--- /dev/null
+++ b/xen/arch/x86/hvm/vmware/guest_msg_def.h
@@ -0,0 +1,87 @@
+/* **********************************************************
+ * Copyright 1998 VMware, Inc. All rights reserved.
+ * **********************************************************
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation version 2 and no later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
+ * for more details.
+ *
+ * You should have received a copy of the GNU General Public License along
+ * with this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/*
+ * guest_msg_def.h --
+ *
+ * Second layer of the internal communication channel between guest
+ * applications and vmware
+ *
+ */
+
+#ifndef _GUEST_MSG_DEF_H_
+#define _GUEST_MSG_DEF_H_
+
+#define INCLUDE_ALLOW_MODULE
+#define INCLUDE_ALLOW_USERLEVEL
+#define INCLUDE_ALLOW_VMMEXT
+#include "includeCheck.h"
+
+
+/* Basic request types */
+typedef enum {
+ MESSAGE_TYPE_OPEN,
+ MESSAGE_TYPE_SENDSIZE,
+ MESSAGE_TYPE_SENDPAYLOAD,
+ MESSAGE_TYPE_RECVSIZE,
+ MESSAGE_TYPE_RECVPAYLOAD,
+ MESSAGE_TYPE_RECVSTATUS,
+ MESSAGE_TYPE_CLOSE,
+} MessageType;
+
+
+/* Reply statuses */
+/* The basic request succeeded */
+#define MESSAGE_STATUS_SUCCESS 0x0001
+/* vmware has a message available for its party */
+#define MESSAGE_STATUS_DORECV 0x0002
+/* The channel has been closed */
+#define MESSAGE_STATUS_CLOSED 0x0004
+/* vmware removed the message before the party fetched it */
+#define MESSAGE_STATUS_UNSENT 0x0008
+/* A checkpoint occurred */
+#define MESSAGE_STATUS_CPT 0x0010
+/* An underlying device is powering off */
+#define MESSAGE_STATUS_POWEROFF 0x0020
+/* vmware has detected a timeout on the channel */
+#define MESSAGE_STATUS_TIMEOUT 0x0040
+/* vmware supports high-bandwidth for sending and receiving the payload */
+#define MESSAGE_STATUS_HB 0x0080
+
+/*
+ * This mask defines the status bits that the guest is allowed to set;
+ * we use this to mask out all other bits when receiving the status
+ * from the guest. Otherwise, the guest can manipulate VMX state by
+ * setting status bits that are only supposed to be changed by the
+ * VMX. See bug 45385.
+ */
+#define MESSAGE_STATUS_GUEST_MASK MESSAGE_STATUS_SUCCESS
+
+/*
+ * Max number of channels.
+ * Unfortunately this has to be public because the monitor part
+ * of the backdoor needs it for its trivial-case optimization. [greg]
+ */
+#define GUESTMSG_MAX_CHANNEL 8
+
+/* Flags to open a channel. --hpreg */
+#define GUESTMSG_FLAG_COOKIE 0x80000000
+#define GUESTMSG_FLAG_ALL GUESTMSG_FLAG_COOKIE
+
+
+#endif /* _GUEST_MSG_DEF_H_ */
diff --git a/xen/arch/x86/hvm/vmware/includeCheck.h b/xen/arch/x86/hvm/vmware/includeCheck.h
new file mode 100644
index 0000000..26e0d59
--- /dev/null
+++ b/xen/arch/x86/hvm/vmware/includeCheck.h
@@ -0,0 +1,17 @@
+/*
+ * includeCheck.h
+ *
+ * Copyright (C) 2012 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+/*
+ * Nothing here. Just to use backdoor_def.h without change.
+ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (2 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 03/16] vmware: Add VMware provided include files Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 21:22 ` Andrew Cooper
2014-09-12 13:08 ` Boris Ostrovsky
2014-09-11 18:36 ` [PATCH v4 05/16] tools: Add vmware_port support Don Slutz
` (11 subsequent siblings)
15 siblings, 2 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
This is a new domain_create() flag, DOMCRF_vmware_port. It is
passed to domctl as XEN_DOMCTL_CDF_vmware_port.
This enables limited support of VMware's hyper-call.
This is both a more complete support then in currently provided by
QEMU and/or KVM and less. The missing part requires QEMU changes
and has been left out until the QEMU patches are accepted upstream.
VMware's hyper-call is also known as VMware Backdoor I/O Port.
Note: this support does not depend on vmware_hw being non-zero.
Summary is that VMware treats "in (%dx),%eax" (or "out %eax,(%dx)")
to port 0x5658 specially. Note: since many operations return data
in EAX, "in (%dx),%eax" is the one to use. The other lengths like
"in (%dx),%al" will still do things, only AL part of EAX will be
changed. For "out %eax,(%dx)" of all lengths, EAX will remain
unchanged.
Also this instruction is allowed to be used from ring 3. To
support this the vmexit for GP needs to be enabled. I have not
fully tested that nested HVM is doing the right thing for this.
An open source example of using this is:
http://open-vm-tools.sourceforge.net/
Which only uses "inl (%dx)". Also
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009458
The support included is enough to allow VMware tools to install in a
HVM domU.
For a debug=y build there is a new command line option
vmport_debug=. It enabled output to the console of various
stages of handling the "in (%dx)" instruction.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
xen/arch/x86/domain.c | 2 +
xen/arch/x86/hvm/hvm.c | 4 +
xen/arch/x86/hvm/svm/emulate.c | 26 ++-
xen/arch/x86/hvm/svm/svm.c | 39 ++++-
xen/arch/x86/hvm/svm/vmcb.c | 2 +
xen/arch/x86/hvm/vmware/Makefile | 1 +
xen/arch/x86/hvm/vmware/vmport.c | 310 ++++++++++++++++++++++++++++++++++
xen/arch/x86/hvm/vmx/vmcs.c | 1 +
xen/arch/x86/hvm/vmx/vmx.c | 60 +++++++
xen/arch/x86/hvm/vmx/vvmx.c | 14 ++
xen/common/domctl.c | 3 +
xen/include/asm-x86/hvm/domain.h | 3 +
xen/include/asm-x86/hvm/io.h | 2 +-
xen/include/asm-x86/hvm/svm/emulate.h | 13 +-
xen/include/asm-x86/hvm/vmport.h | 61 +++++++
xen/include/public/domctl.h | 3 +
xen/include/xen/sched.h | 3 +
17 files changed, 536 insertions(+), 11 deletions(-)
create mode 100644 xen/arch/x86/hvm/vmware/vmport.c
create mode 100644 xen/include/asm-x86/hvm/vmport.h
diff --git a/xen/arch/x86/domain.c b/xen/arch/x86/domain.c
index f7e0e78..a6b82de 100644
--- a/xen/arch/x86/domain.c
+++ b/xen/arch/x86/domain.c
@@ -507,6 +507,8 @@ int arch_domain_create(struct domain *d, unsigned int domcr_flags)
d->arch.hvm_domain.mem_sharing_enabled = 0;
d->arch.s3_integrity = !!(domcr_flags & DOMCRF_s3_integrity);
+ d->arch.hvm_domain.is_vmware_port_enabled =
+ (domcr_flags & DOMCRF_vmware_port);
INIT_LIST_HEAD(&d->arch.pdev_list);
diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 03a1a19..9c59c85 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -58,6 +58,7 @@
#include <asm/hvm/trace.h>
#include <asm/hvm/nestedhvm.h>
#include <asm/hvm/vmware.h>
+#include <asm/hvm/vmport.h>
#include <asm/mtrr.h>
#include <asm/apic.h>
#include <public/sched.h>
@@ -1498,6 +1499,9 @@ int hvm_domain_initialise(struct domain *d)
goto fail1;
d->arch.hvm_domain.io_handler->num_slot = 0;
+ if ( d->arch.hvm_domain.is_vmware_port_enabled )
+ vmport_register(d);
+
if ( is_pvh_domain(d) )
{
register_portio_handler(d, 0, 0x10003, handle_pvh_io);
diff --git a/xen/arch/x86/hvm/svm/emulate.c b/xen/arch/x86/hvm/svm/emulate.c
index 37a1ece..2446eb7 100644
--- a/xen/arch/x86/hvm/svm/emulate.c
+++ b/xen/arch/x86/hvm/svm/emulate.c
@@ -50,7 +50,7 @@ static unsigned int is_prefix(u8 opc)
return 0;
}
-static unsigned long svm_rip2pointer(struct vcpu *v)
+unsigned long svm_rip2pointer(struct vcpu *v)
{
struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
unsigned long p = vmcb->cs.base + guest_cpu_user_regs()->eip;
@@ -78,6 +78,7 @@ static unsigned long svm_nextrip_insn_length(struct vcpu *v)
/* ...and the rest of the #VMEXITs */
case VMEXIT_CR0_SEL_WRITE:
case VMEXIT_EXCEPTION_BP:
+ case VMEXIT_EXCEPTION_GP:
break;
default:
BUG();
@@ -95,6 +96,10 @@ MAKE_INSTR(CPUID, 2, 0x0f, 0xa2);
MAKE_INSTR(RDMSR, 2, 0x0f, 0x32);
MAKE_INSTR(WRMSR, 2, 0x0f, 0x30);
MAKE_INSTR(VMCALL, 3, 0x0f, 0x01, 0xd9);
+MAKE_INSTR(INB_DX, 1, 0xec);
+MAKE_INSTR(INL_DX, 1, 0xed);
+MAKE_INSTR(OUTB_DX,1, 0xee);
+MAKE_INSTR(OUTL_DX,1, 0xef);
MAKE_INSTR(HLT, 1, 0xf4);
MAKE_INSTR(INT3, 1, 0xcc);
MAKE_INSTR(RDTSC, 2, 0x0f, 0x31);
@@ -115,6 +120,10 @@ static const u8 *const opc_bytes[INSTR_MAX_COUNT] =
[INSTR_RDMSR] = OPCODE_RDMSR,
[INSTR_WRMSR] = OPCODE_WRMSR,
[INSTR_VMCALL] = OPCODE_VMCALL,
+ [INSTR_INB_DX] = OPCODE_INB_DX,
+ [INSTR_INL_DX] = OPCODE_INL_DX,
+ [INSTR_OUTB_DX] = OPCODE_OUTB_DX,
+ [INSTR_OUTL_DX] = OPCODE_OUTL_DX,
[INSTR_HLT] = OPCODE_HLT,
[INSTR_INT3] = OPCODE_INT3,
[INSTR_RDTSC] = OPCODE_RDTSC,
@@ -152,7 +161,9 @@ static int fetch(struct vcpu *v, u8 *buf, unsigned long addr, int len)
}
int __get_instruction_length_from_list(struct vcpu *v,
- const enum instruction_index *list, unsigned int list_count)
+ const enum instruction_index *list,
+ unsigned int list_count,
+ bool_t err_rpt)
{
struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
unsigned int i, j, inst_len = 0;
@@ -211,10 +222,13 @@ int __get_instruction_length_from_list(struct vcpu *v,
mismatch: ;
}
- gdprintk(XENLOG_WARNING,
- "%s: Mismatch between expected and actual instruction bytes: "
- "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
- hvm_inject_hw_exception(TRAP_gp_fault, 0);
+ if ( err_rpt )
+ {
+ gdprintk(XENLOG_WARNING,
+ "%s: Mismatch between expected and actual instruction bytes: "
+ "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
+ hvm_inject_hw_exception(TRAP_gp_fault, 0);
+ }
return 0;
done:
diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
index b5188e6..9e14d2a 100644
--- a/xen/arch/x86/hvm/svm/svm.c
+++ b/xen/arch/x86/hvm/svm/svm.c
@@ -59,6 +59,7 @@
#include <public/sched.h>
#include <asm/hvm/vpt.h>
#include <asm/hvm/trace.h>
+#include <asm/hvm/vmport.h>
#include <asm/hap.h>
#include <asm/apic.h>
#include <asm/debugger.h>
@@ -2065,6 +2066,38 @@ svm_vmexit_do_vmsave(struct vmcb_struct *vmcb,
return;
}
+static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
+ struct vcpu *v)
+{
+ struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
+ unsigned long inst_len;
+ unsigned long inst_addr = svm_rip2pointer(v);
+ int rc;
+ static const enum instruction_index list[] = {
+ INSTR_INL_DX, INSTR_INB_DX, INSTR_OUTL_DX, INSTR_OUTB_DX
+ };
+
+ inst_len = __get_instruction_length_from_list(
+ v, list, ARRAY_SIZE(list), 0);
+
+ rc = vmport_gp_check(regs, v, inst_len, inst_addr,
+ vmcb->exitinfo1, vmcb->exitinfo2);
+ if ( !rc )
+ __update_guest_eip(regs, inst_len);
+ else
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_GP_UNKNOWN,
+ "gp: rc=%d ei1=0x%lx ei2=0x%lx ip=%"PRIx64
+ " (0x%lx,%ld) ax=%"PRIx64" bx=%"PRIx64" cx=%"PRIx64
+ " dx=%"PRIx64" si=%"PRIx64" di=%"PRIx64, rc,
+ (unsigned long)vmcb->exitinfo1,
+ (unsigned long)vmcb->exitinfo2, regs->rip, inst_addr,
+ inst_len, regs->rax, regs->rbx, regs->rcx, regs->rdx,
+ regs->rsi, regs->rdi);
+ hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
+ }
+}
+
static void svm_vmexit_ud_intercept(struct cpu_user_regs *regs)
{
struct hvm_emulate_ctxt ctxt;
@@ -2140,7 +2173,7 @@ static void svm_vmexit_do_invalidate_cache(struct cpu_user_regs *regs)
int inst_len;
inst_len = __get_instruction_length_from_list(
- current, list, ARRAY_SIZE(list));
+ current, list, ARRAY_SIZE(list), 1);
if ( inst_len == 0 )
return;
@@ -2412,6 +2445,10 @@ void svm_vmexit_handler(struct cpu_user_regs *regs)
break;
}
+ case VMEXIT_EXCEPTION_GP:
+ svm_vmexit_gp_intercept(regs, v);
+ break;
+
case VMEXIT_EXCEPTION_UD:
svm_vmexit_ud_intercept(regs);
break;
diff --git a/xen/arch/x86/hvm/svm/vmcb.c b/xen/arch/x86/hvm/svm/vmcb.c
index 21292bb..914d1ad 100644
--- a/xen/arch/x86/hvm/svm/vmcb.c
+++ b/xen/arch/x86/hvm/svm/vmcb.c
@@ -195,6 +195,8 @@ static int construct_vmcb(struct vcpu *v)
HVM_TRAP_MASK
| (1U << TRAP_no_device);
+ if ( v->domain->arch.hvm_domain.is_vmware_port_enabled )
+ vmcb->_exception_intercepts |= TRAP_gp_fault;
if ( paging_mode_hap(v->domain) )
{
vmcb->_np_enable = 1; /* enable nested paging */
diff --git a/xen/arch/x86/hvm/vmware/Makefile b/xen/arch/x86/hvm/vmware/Makefile
index 3fb2e0b..cd8815b 100644
--- a/xen/arch/x86/hvm/vmware/Makefile
+++ b/xen/arch/x86/hvm/vmware/Makefile
@@ -1 +1,2 @@
obj-y += cpuid.o
+obj-y += vmport.o
diff --git a/xen/arch/x86/hvm/vmware/vmport.c b/xen/arch/x86/hvm/vmware/vmport.c
new file mode 100644
index 0000000..bfc01dc
--- /dev/null
+++ b/xen/arch/x86/hvm/vmware/vmport.c
@@ -0,0 +1,310 @@
+/*
+ * HVM VMPORT emulation
+ *
+ * Copyright (C) 2012 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#include <xen/config.h>
+#include <xen/lib.h>
+#include <asm/hvm/hvm.h>
+#include <asm/hvm/support.h>
+#include <asm/hvm/vmport.h>
+
+#include "backdoor_def.h"
+#include "guest_msg_def.h"
+
+#ifndef NDEBUG
+unsigned int opt_vmport_debug __read_mostly;
+integer_param("vmport_debug", opt_vmport_debug);
+#endif
+
+/* More VMware defines */
+
+#define VMWARE_GUI_AUTO_GRAB 0x001
+#define VMWARE_GUI_AUTO_UNGRAB 0x002
+#define VMWARE_GUI_AUTO_SCROLL 0x004
+#define VMWARE_GUI_AUTO_RAISE 0x008
+#define VMWARE_GUI_EXCHANGE_SELECTIONS 0x010
+#define VMWARE_GUI_WARP_CURSOR_ON_UNGRAB 0x020
+#define VMWARE_GUI_FULL_SCREEN 0x040
+
+#define VMWARE_GUI_TO_FULL_SCREEN 0x080
+#define VMWARE_GUI_TO_WINDOW 0x100
+
+#define VMWARE_GUI_AUTO_RAISE_DISABLED 0x200
+
+#define VMWARE_GUI_SYNC_TIME 0x400
+
+/* When set, toolboxes should not show the cursor options page. */
+#define VMWARE_DISABLE_CURSOR_OPTIONS 0x800
+
+inline uint16_t get_low_bits(uint32_t bits)
+{
+ return bits & 0xffff;
+}
+
+void vmport_register(struct domain *d)
+{
+ register_portio_handler(d, BDOOR_PORT, 4, vmport_ioport);
+}
+
+int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val)
+{
+ struct cpu_user_regs *regs = guest_cpu_user_regs();
+ uint32_t cmd = get_low_bits(regs->rcx);
+ uint32_t magic = regs->rax;
+ int rc = X86EMUL_OKAY;
+
+ if ( magic == BDOOR_MAGIC )
+ {
+ uint64_t saved_rax = regs->rax;
+ uint64_t value;
+
+ VMPORT_DBG_LOG(VMPORT_LOG_TRACE,
+ "VMware trace dir=%d bytes=%u ip=%"PRIx64" cmd=%d ax=%"
+ PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64" si=%"
+ PRIx64" di=%"PRIx64"\n", dir, bytes,
+ regs->rip, cmd, regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
+ switch ( cmd )
+ {
+ case BDOOR_CMD_GETMHZ:
+ /* ... */
+ regs->rbx = BDOOR_MAGIC;
+ regs->rax = current->domain->arch.tsc_khz / 1000;
+ break;
+ case BDOOR_CMD_GETVERSION:
+ /* ... */
+ regs->rbx = BDOOR_MAGIC;
+ /* VERSION_MAGIC */
+ regs->rax = 6;
+ /* Claim we are an ESX. VMX_TYPE_SCALABLE_SERVER */
+ regs->rcx = 2;
+ break;
+ case BDOOR_CMD_GETHWVERSION:
+ /* ... */
+ regs->rbx = BDOOR_MAGIC;
+ /* vmware_hw */
+ regs->rax = 0;
+ if ( is_hvm_vcpu(current) )
+ {
+ struct hvm_domain *hd = ¤t->domain->arch.hvm_domain;
+
+ regs->rax = hd->params[HVM_PARAM_VMWARE_HW];
+ }
+ if ( !regs->rax )
+ regs->rax = 4; /* Act like version 4 */
+ break;
+ case BDOOR_CMD_GETHZ:
+ value = current->domain->arch.tsc_khz * 1000;
+ /* apic-frequency (bus speed) */
+ regs->rcx = (uint32_t)(1000000000ULL / APIC_BUS_CYCLE_NS);
+ /* High part of tsc-frequency */
+ regs->rbx = (uint32_t)(value >> 32);
+ /* Low part of tsc-frequency */
+ regs->rax = value;
+ break;
+ case BDOOR_CMD_GETTIME:
+ value = get_localtime_us(current->domain);
+ /* hostUsecs */
+ regs->rbx = (uint32_t)(value % 1000000UL);
+ /* hostSecs */
+ regs->rax = value / 1000000ULL;
+ /* maxTimeLag */
+ regs->rcx = 0;
+ break;
+ case BDOOR_CMD_GETTIMEFULL:
+ value = get_localtime_us(current->domain);
+ /* ... */
+ regs->rax = BDOOR_MAGIC;
+ /* hostUsecs */
+ regs->rbx = (uint32_t)(value % 1000000UL);
+ /* High part of hostSecs */
+ regs->rsi = (uint32_t)((value / 1000000ULL) >> 32);
+ /* Low part of hostSecs */
+ regs->rdx = (uint32_t)(value / 1000000ULL);
+ /* maxTimeLag */
+ regs->rcx = 0;
+ break;
+ case BDOOR_CMD_GETGUIOPTIONS:
+ regs->rax = VMWARE_GUI_AUTO_GRAB | VMWARE_GUI_AUTO_UNGRAB |
+ VMWARE_GUI_AUTO_RAISE_DISABLED | VMWARE_GUI_SYNC_TIME |
+ VMWARE_DISABLE_CURSOR_OPTIONS;
+ break;
+ case BDOOR_CMD_SETGUIOPTIONS:
+ regs->rax = 0x0;
+ break;
+ default:
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
+ "VMware bytes=%d dir=%d cmd=%d",
+ bytes, dir, cmd);
+ break;
+ }
+ VMPORT_DBG_LOG(VMPORT_LOG_VMWARE_AFTER,
+ "VMware after ip=%"PRIx64" cmd=%d ax=%"PRIx64" bx=%"
+ PRIx64" cx=%"PRIx64" dx=%"PRIx64" si=%"PRIx64" di=%"
+ PRIx64"\n",
+ regs->rip, cmd, regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
+ if ( dir == IOREQ_READ )
+ {
+ switch ( bytes )
+ {
+ case 1:
+ regs->rax = (saved_rax & 0xffffff00) | (regs->rax & 0xff);
+ break;
+ case 2:
+ regs->rax = (saved_rax & 0xffff0000) | get_low_bits(regs->rax);
+ break;
+ case 4:
+ regs->rax = (uint32_t)regs->rax;
+ break;
+ }
+ *val = regs->rax;
+ }
+ else
+ regs->rax = saved_rax;
+ }
+ else
+ {
+ rc = X86EMUL_UNHANDLEABLE;
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
+ "Not VMware %x vs %x; ip=%"PRIx64" ax=%"PRIx64
+ " bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64" si=%"PRIx64
+ " di=%"PRIx64"",
+ magic, BDOOR_MAGIC, regs->rip, regs->rax, regs->rbx,
+ regs->rcx, regs->rdx, regs->rsi, regs->rdi);
+ }
+
+ return rc;
+}
+
+int vmport_gp_check(struct cpu_user_regs *regs, struct vcpu *v,
+ unsigned long inst_len, unsigned long inst_addr,
+ unsigned long ei1, unsigned long ei2)
+{
+ if ( !v->domain->arch.hvm_domain.is_vmware_port_enabled )
+ return 10;
+
+ if ( inst_len && inst_len <= 2 && get_low_bits(regs->rdx) == BDOOR_PORT &&
+ ei1 == 0 && ei2 == 0 && regs->error_code == 0 &&
+ (uint32_t)regs->rax == BDOOR_MAGIC )
+ {
+ int i = 0;
+ uint32_t val;
+ uint32_t byte_cnt = 4;
+ unsigned char bytes[2];
+ unsigned int fetch_len;
+ int frc;
+ int rc;
+
+ /*
+ * Fetch up to the next page break; we'll fetch from the
+ * next page later if we have to.
+ */
+ fetch_len = min_t(unsigned int, inst_len,
+ PAGE_SIZE - (inst_addr & ~PAGE_MASK));
+ frc = hvm_fetch_from_guest_virt_nofault(bytes, inst_addr, fetch_len,
+ PFEC_page_present);
+ if ( frc != HVMCOPY_okay )
+ {
+ gdprintk(XENLOG_WARNING,
+ "Bad instruction fetch at %#lx (frc=%d il=%lu fl=%u)\n",
+ (unsigned long) inst_addr, frc, inst_len, fetch_len);
+ return 11;
+ }
+ if ( bytes[0] == 0x66 ) /* operand size prefix */
+ {
+ byte_cnt = 2;
+ i = 1;
+ if ( fetch_len != inst_len )
+ {
+ frc = hvm_fetch_from_guest_virt_nofault(&bytes[1],
+ inst_addr + 1, 1,
+ PFEC_page_present);
+ if ( frc != HVMCOPY_okay )
+ {
+ gdprintk(XENLOG_WARNING,
+ "Bad instruction fetch at %#lx + 1 (frc=%d)\n",
+ (unsigned long) inst_addr, frc);
+ return 12;
+ }
+ }
+ }
+ if ( bytes[i] == 0xed ) /* in (%dx),%eax or in (%dx),%ax */
+ {
+ rc = vmport_ioport(IOREQ_READ, BDOOR_PORT, byte_cnt, &val);
+ VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
+ "gp: VMwareIn rc=%d ip=%"PRIx64" byte_cnt=%d ax=%"
+ PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
+ " si=%"PRIx64" di=%"PRIx64, rc,
+ inst_addr, byte_cnt, regs->rax, regs->rbx,
+ regs->rcx, regs->rdx, regs->rsi, regs->rdi);
+ return rc;
+ }
+ else if ( bytes[i] == 0xec ) /* in (%dx),%al */
+ {
+ rc = vmport_ioport(IOREQ_READ, BDOOR_PORT, 1, &val);
+ VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
+ "gp: VMwareIn rc=%d ip=%"PRIx64" byte_cnt=1 ax=%"
+ PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
+ " si=%"PRIx64" di=%"PRIx64, rc,
+ inst_addr, regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
+ return rc;
+ }
+ else if ( bytes[i] == 0xef ) /* out %eax,(%dx) or out %ax,(%dx) */
+ {
+ rc = vmport_ioport(IOREQ_WRITE, BDOOR_PORT, byte_cnt, &val);
+ VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
+ "gp: VMwareOut rc=%d ip=%"PRIx64" byte_cnt=%d ax=%"
+ PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
+ " si=%"PRIx64" di=%"PRIx64, rc,
+ inst_addr, byte_cnt, regs->rax, regs->rbx,
+ regs->rcx, regs->rdx, regs->rsi, regs->rdi);
+ return rc;
+ }
+ else if ( bytes[i] == 0xee ) /* out %al,(%dx) */
+ {
+ rc = vmport_ioport(IOREQ_WRITE, BDOOR_PORT, 1, &val);
+ VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
+ "gp: VMwareOut rc=%d ip=%"PRIx64" byte_cnt=1 ax=%"
+ PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
+ " si=%"PRIx64" di=%"PRIx64, rc,
+ inst_addr, regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
+ return rc;
+ }
+ else
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_GP_FAIL_RD_INST,
+ "gp: VMware? lip=%"PRIx64"[%d]=>0x%x(%ld) ax=%"
+ PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
+ " si=%"PRIx64" di=%"PRIx64,
+ inst_addr, i, bytes[i], inst_len, regs->rax,
+ regs->rbx, regs->rcx, regs->rdx, regs->rsi,
+ regs->rdi);
+ return 13;
+ }
+ }
+ return 14;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/arch/x86/hvm/vmx/vmcs.c b/xen/arch/x86/hvm/vmx/vmcs.c
index 4a4f4e1..fa1f69a 100644
--- a/xen/arch/x86/hvm/vmx/vmcs.c
+++ b/xen/arch/x86/hvm/vmx/vmcs.c
@@ -1077,6 +1077,7 @@ static int construct_vmcs(struct vcpu *v)
v->arch.hvm_vmx.exception_bitmap = HVM_TRAP_MASK
| (paging_mode_hap(d) ? 0 : (1U << TRAP_page_fault))
+ | (1U << TRAP_gp_fault)
| (1U << TRAP_no_device);
vmx_update_exception_bitmap(v);
diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
index 26b1ad5..6cd1730 100644
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -44,6 +44,7 @@
#include <asm/hvm/support.h>
#include <asm/hvm/vmx/vmx.h>
#include <asm/hvm/vmx/vmcs.h>
+#include <asm/hvm/vmport.h>
#include <public/sched.h>
#include <public/hvm/ioreq.h>
#include <asm/hvm/vpic.h>
@@ -1279,6 +1280,8 @@ static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr)
v->arch.hvm_vmx.exception_bitmap = HVM_TRAP_MASK
| (paging_mode_hap(v->domain) ?
0 : (1U << TRAP_page_fault))
+ | v->domain->arch.hvm_domain.is_vmware_port_enabled ?
+ (1U << TRAP_gp_fault) : 0
| (1U << TRAP_no_device);
vmx_update_exception_bitmap(v);
vmx_update_debug_state(v);
@@ -2565,6 +2568,49 @@ static void vmx_idtv_reinject(unsigned long idtv_info)
}
}
+static unsigned long vmx_rip2pointer(struct cpu_user_regs *regs,
+ struct vcpu *v)
+{
+ struct segment_register cs;
+ unsigned long p;
+
+ vmx_get_segment_register(v, x86_seg_cs, &cs);
+ p = cs.base + regs->rip;
+ if ( !(cs.attr.fields.l && hvm_long_mode_enabled(v)) )
+ return (uint32_t)p; /* mask to 32 bits */
+ return p;
+}
+
+static void vmx_vmexit_gp_intercept(struct cpu_user_regs *regs,
+ struct vcpu *v)
+{
+ unsigned long exit_qualification;
+ unsigned long inst_len;
+ unsigned long inst_addr = vmx_rip2pointer(regs, v);
+ unsigned long ecode;
+ int rc;
+
+ __vmread(EXIT_QUALIFICATION, &exit_qualification);
+ __vmread(VM_EXIT_INSTRUCTION_LEN, &inst_len);
+ __vmread(VM_EXIT_INTR_ERROR_CODE, &ecode);
+
+ rc = vmport_gp_check(regs, v, inst_len, inst_addr,
+ ecode, exit_qualification);
+ if ( !rc )
+ update_guest_eip();
+ else
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_GP_UNKNOWN,
+ "gp: rc=%d ecode=0x%lx eq=0x%lx ip=%"PRIx64
+ " (0x%lx,%ld) ax=%"PRIx64" bx=%"PRIx64" cx=%"PRIx64
+ " dx=%"PRIx64" si=%"PRIx64" di=%"PRIx64, rc, ecode,
+ exit_qualification, regs->rip, inst_addr, inst_len,
+ regs->rax, regs->rbx, regs->rcx, regs->rdx, regs->rsi,
+ regs->rdi);
+ hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
+ }
+}
+
static int vmx_handle_apic_write(void)
{
unsigned long exit_qualification;
@@ -2675,6 +2721,17 @@ void vmx_vmexit_handler(struct cpu_user_regs *regs)
&& vector != TRAP_nmi
&& vector != TRAP_machine_check )
{
+#ifndef NDEBUG
+ if ( vector == TRAP_gp_fault )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_REALMODE_GP,
+ "realmode gp: ip=%"PRIx64" ax=%"PRIx64
+ " bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
+ " si=%"PRIx64" di=%"PRIx64,
+ regs->rip, regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
+ }
+#endif
perfc_incr(realmode_exits);
v->arch.hvm_vmx.vmx_emulate = 1;
HVMTRACE_0D(REALMODE_EMULATE);
@@ -2790,6 +2847,9 @@ void vmx_vmexit_handler(struct cpu_user_regs *regs)
HVMTRACE_1D(TRAP, vector);
vmx_fpu_dirty_intercept();
break;
+ case TRAP_gp_fault:
+ vmx_vmexit_gp_intercept(regs, v);
+ break;
case TRAP_page_fault:
__vmread(EXIT_QUALIFICATION, &exit_qualification);
__vmread(VM_EXIT_INTR_ERROR_CODE, &ecode);
diff --git a/xen/arch/x86/hvm/vmx/vvmx.c b/xen/arch/x86/hvm/vmx/vvmx.c
index 9ccc03f..51d2336 100644
--- a/xen/arch/x86/hvm/vmx/vvmx.c
+++ b/xen/arch/x86/hvm/vmx/vvmx.c
@@ -24,6 +24,7 @@
#include <asm/types.h>
#include <asm/mtrr.h>
#include <asm/p2m.h>
+#include <asm/hvm/vmport.h>
#include <asm/hvm/vmx/vmx.h>
#include <asm/hvm/vmx/vvmx.h>
#include <asm/hvm/nestedhvm.h>
@@ -2182,6 +2183,19 @@ int nvmx_n2_vmexit_handler(struct cpu_user_regs *regs,
if ( v->fpu_dirtied )
nvcpu->nv_vmexit_pending = 1;
}
+ else if ( vector == TRAP_gp_fault )
+ {
+#ifndef NDEBUG
+ struct cpu_user_regs *ur = guest_cpu_user_regs();
+ VMPORT_DBG_LOG(VMPORT_LOG_VGP_UNKNOWN,
+ "Unexpected gp: ip=%"PRIx64" ax=%"PRIx64
+ " bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
+ " si=%"PRIx64" di=%"PRIx64,
+ ur->rip, ur->rax, ur->rbx, ur->rcx, ur->rdx,
+ ur->rsi, ur->rdi);
+#endif
+ nvcpu->nv_vmexit_pending = 1;
+ }
else if ( (intr_info & valid_mask) == valid_mask )
{
exec_bitmap =__get_vvmcs(nvcpu->nv_vvmcx, EXCEPTION_BITMAP);
diff --git a/xen/common/domctl.c b/xen/common/domctl.c
index 8907aac..1307be0 100644
--- a/xen/common/domctl.c
+++ b/xen/common/domctl.c
@@ -541,6 +541,7 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
~(XEN_DOMCTL_CDF_hvm_guest
| XEN_DOMCTL_CDF_pvh_guest
| XEN_DOMCTL_CDF_hap
+ | XEN_DOMCTL_CDF_vmware_port
| XEN_DOMCTL_CDF_s3_integrity
| XEN_DOMCTL_CDF_oos_off)) )
break;
@@ -584,6 +585,8 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl)
domcr_flags |= DOMCRF_s3_integrity;
if ( op->u.createdomain.flags & XEN_DOMCTL_CDF_oos_off )
domcr_flags |= DOMCRF_oos_off;
+ if ( op->u.createdomain.flags & XEN_DOMCTL_CDF_vmware_port )
+ domcr_flags |= DOMCRF_vmware_port;
d = domain_create(dom, domcr_flags, op->u.createdomain.ssidref);
if ( IS_ERR(d) )
diff --git a/xen/include/asm-x86/hvm/domain.h b/xen/include/asm-x86/hvm/domain.h
index 291a2e0..93b081b 100644
--- a/xen/include/asm-x86/hvm/domain.h
+++ b/xen/include/asm-x86/hvm/domain.h
@@ -121,6 +121,9 @@ struct hvm_domain {
spinlock_t uc_lock;
bool_t is_in_uc_mode;
+ /* VMware backdoor port available */
+ bool_t is_vmware_port_enabled;
+
/* Pass-through */
struct hvm_iommu hvm_iommu;
diff --git a/xen/include/asm-x86/hvm/io.h b/xen/include/asm-x86/hvm/io.h
index 886a9d6..d257161 100644
--- a/xen/include/asm-x86/hvm/io.h
+++ b/xen/include/asm-x86/hvm/io.h
@@ -25,7 +25,7 @@
#include <public/hvm/ioreq.h>
#include <public/event_channel.h>
-#define MAX_IO_HANDLER 16
+#define MAX_IO_HANDLER 17
#define HVM_PORTIO 0
#define HVM_BUFFERED_IO 2
diff --git a/xen/include/asm-x86/hvm/svm/emulate.h b/xen/include/asm-x86/hvm/svm/emulate.h
index ccc2d3c..542b6b8 100644
--- a/xen/include/asm-x86/hvm/svm/emulate.h
+++ b/xen/include/asm-x86/hvm/svm/emulate.h
@@ -39,18 +39,25 @@ enum instruction_index {
INSTR_STGI,
INSTR_CLGI,
INSTR_INVLPGA,
+ INSTR_INL_DX,
+ INSTR_INB_DX,
+ INSTR_OUTL_DX,
+ INSTR_OUTB_DX,
INSTR_MAX_COUNT /* Must be last - Number of instructions supported */
};
struct vcpu;
-int __get_instruction_length_from_list(
- struct vcpu *, const enum instruction_index *, unsigned int list_count);
+unsigned long svm_rip2pointer(struct vcpu *v);
+int __get_instruction_length_from_list(struct vcpu *,
+ const enum instruction_index *,
+ unsigned int list_count,
+ bool_t err_rpt);
static inline int __get_instruction_length(
struct vcpu *v, enum instruction_index instr)
{
- return __get_instruction_length_from_list(v, &instr, 1);
+ return __get_instruction_length_from_list(v, &instr, 1, 1);
}
#endif /* __ASM_X86_HVM_SVM_EMULATE_H__ */
diff --git a/xen/include/asm-x86/hvm/vmport.h b/xen/include/asm-x86/hvm/vmport.h
new file mode 100644
index 0000000..ea2c5a9
--- /dev/null
+++ b/xen/include/asm-x86/hvm/vmport.h
@@ -0,0 +1,61 @@
+/*
+ * asm/hvm/vmport.h: HVM VMPORT emulation
+ *
+ *
+ * Copyright (C) 2012 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef ASM_X86_HVM_VMPORT_H__
+#define ASM_X86_HVM_VMPORT_H__
+
+#ifndef NDEBUG
+
+#define VMPORT_LOG_GP_UNKNOWN (1 << 0)
+#define VMPORT_LOG_GP_VMWARE_AFTER (1 << 1)
+#define VMPORT_LOG_GP_FAIL_RD_INST (1 << 2)
+#define VMPORT_LOG_VGP_UNKNOWN (1 << 3)
+#define VMPORT_LOG_REALMODE_GP (1 << 4)
+
+#define VMPORT_LOG_GP_NOT_VMWARE (1 << 9)
+
+#define VMPORT_LOG_TRACE (1 << 16)
+#define VMPORT_LOG_ERROR (1 << 17)
+#define VMPORT_LOG_VMWARE_AFTER (1 << 18)
+
+extern unsigned int opt_vmport_debug;
+#define VMPORT_DBG_LOG(level, _f, _a...) \
+ do { \
+ if ( unlikely((level) & opt_vmport_debug) ) \
+ printk("[HVM:%d.%d] <%s> " _f "\n", \
+ current->domain->domain_id, current->vcpu_id, __func__, \
+ ## _a); \
+ } while ( 0 )
+#else
+#define VMPORT_DBG_LOG(level, _f, _a...) do {} while ( 0 )
+#endif
+
+void vmport_register(struct domain *d);
+int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val);
+int vmport_gp_check(struct cpu_user_regs *regs, struct vcpu *v,
+ unsigned long inst_len, unsigned long inst_addr,
+ unsigned long ei1, unsigned long ei2);
+
+#endif /* ASM_X86_HVM_VMPORT_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h
index 69a8b44..7a0f691 100644
--- a/xen/include/public/domctl.h
+++ b/xen/include/public/domctl.h
@@ -63,6 +63,9 @@ struct xen_domctl_createdomain {
/* Is this a PVH guest (as opposed to an HVM or PV guest)? */
#define _XEN_DOMCTL_CDF_pvh_guest 4
#define XEN_DOMCTL_CDF_pvh_guest (1U<<_XEN_DOMCTL_CDF_pvh_guest)
+ /* Is VMware backdoor port available? */
+#define _XEN_DOMCTL_CDF_vmware_port 5
+#define XEN_DOMCTL_CDF_vmware_port (1U<<_XEN_DOMCTL_CDF_vmware_port)
uint32_t flags;
};
typedef struct xen_domctl_createdomain xen_domctl_createdomain_t;
diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
index c5157e6..d741978 100644
--- a/xen/include/xen/sched.h
+++ b/xen/include/xen/sched.h
@@ -546,6 +546,9 @@ struct domain *domain_create(
/* DOMCRF_pvh: Create PV domain in HVM container. */
#define _DOMCRF_pvh 5
#define DOMCRF_pvh (1U<<_DOMCRF_pvh)
+ /* DOMCRF_vmware_port: Enable use of vmware backdoor port. */
+#define _DOMCRF_vmware_port 6
+#define DOMCRF_vmware_port (1U<<_DOMCRF_vmware_port)
/*
* rcu_lock_domain_by_id() is more efficient than get_domain_by_id().
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 05/16] tools: Add vmware_port support
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (3 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 04/16] xen: Add is_vmware_port_enabled Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 06/16] xen: Convert vmware_port to xentrace usage Don Slutz
` (10 subsequent siblings)
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
This new libxl_domain_create_info field is used to set
XEN_DOMCTL_CDF_vmware_port for the xc_domain_create() routine.
In xen it is is_vmware_port_enabled.
If is_vmware_port_enabled then
enable a limited support of VMware's hyper-call.
VMware's hyper-call is also known as VMware Backdoor I/O Port.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
docs/man/xl.cfg.pod.5 | 7 +++++++
tools/libxl/libxl.h | 5 +++++
tools/libxl/libxl_create.c | 2 ++
tools/libxl/libxl_types.idl | 1 +
tools/libxl/xl_cmdimpl.c | 1 +
5 files changed, 16 insertions(+)
diff --git a/docs/man/xl.cfg.pod.5 b/docs/man/xl.cfg.pod.5
index 367b401..ab645d8 100644
--- a/docs/man/xl.cfg.pod.5
+++ b/docs/man/xl.cfg.pod.5
@@ -1164,6 +1164,13 @@ For vssd:VirtualSystemType == vmx-07, vmware_hw = 7.
=back
+=item B<vmware_port=BOOLEAN>
+
+Turns on or off the exposure of VMware port. This is known as
+vmport in QEMU. Also called VMware Backdoor I/O Port. Not all
+defined VMware backdoor commands are implemented. All of the
+ones that Linux kernel uses are defined.
+
=back
=head3 Emulated VGA Graphics Device
diff --git a/tools/libxl/libxl.h b/tools/libxl/libxl.h
index 067af5b..61b7968 100644
--- a/tools/libxl/libxl.h
+++ b/tools/libxl/libxl.h
@@ -148,6 +148,11 @@
#define LIBXL_HAVE_BUILDINFO_HVM_VMWARE_HW 1
/*
+ * libxl_domain_create_info has the vmware_port field.
+ */
+#define LIBXL_HAVE_CREATEINFO_VMWARE_PORT 1
+
+/*
* libxl ABI compatibility
*
* The only guarantee which libxl makes regarding ABI compatibility
diff --git a/tools/libxl/libxl_create.c b/tools/libxl/libxl_create.c
index d9b60af..8cb510f 100644
--- a/tools/libxl/libxl_create.c
+++ b/tools/libxl/libxl_create.c
@@ -38,6 +38,7 @@ int libxl__domain_create_info_setdefault(libxl__gc *gc,
libxl_defbool_setdefault(&c_info->hap, libxl_defbool_val(c_info->pvh));
}
+ libxl_defbool_setdefault(&c_info->vmware_port, false);
libxl_defbool_setdefault(&c_info->run_hotplug_scripts, true);
libxl_defbool_setdefault(&c_info->driver_domain, false);
@@ -501,6 +502,7 @@ int libxl__domain_make(libxl__gc *gc, libxl_domain_create_info *info,
flags |= XEN_DOMCTL_CDF_hvm_guest;
flags |= libxl_defbool_val(info->hap) ? XEN_DOMCTL_CDF_hap : 0;
flags |= libxl_defbool_val(info->oos) ? 0 : XEN_DOMCTL_CDF_oos_off;
+ flags |= libxl_defbool_val(info->vmware_port)? XEN_DOMCTL_CDF_vmware_port : 0;
} else if (libxl_defbool_val(info->pvh)) {
flags |= XEN_DOMCTL_CDF_pvh_guest;
if (!libxl_defbool_val(info->hap)) {
diff --git a/tools/libxl/libxl_types.idl b/tools/libxl/libxl_types.idl
index 907572c..608b64d 100644
--- a/tools/libxl/libxl_types.idl
+++ b/tools/libxl/libxl_types.idl
@@ -294,6 +294,7 @@ libxl_domain_create_info = Struct("domain_create_info",[
("type", libxl_domain_type),
("hap", libxl_defbool),
("oos", libxl_defbool),
+ ("vmware_port", libxl_defbool),
("ssidref", uint32),
("ssid_label", string),
("name", string),
diff --git a/tools/libxl/xl_cmdimpl.c b/tools/libxl/xl_cmdimpl.c
index db86382..549d3db 100644
--- a/tools/libxl/xl_cmdimpl.c
+++ b/tools/libxl/xl_cmdimpl.c
@@ -866,6 +866,7 @@ static void parse_config_data(const char *config_source,
}
xlu_cfg_get_defbool(config, "oos", &c_info->oos, 0);
+ xlu_cfg_get_defbool(config, "vmware_port", &c_info->vmware_port, 0);
if (!xlu_cfg_get_string (config, "pool", &buf, 0))
xlu_cfg_replace_string(config, "pool", &c_info->pool_name, 0);
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 06/16] xen: Convert vmware_port to xentrace usage
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (4 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 05/16] tools: Add vmware_port support Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-12 13:10 ` Boris Ostrovsky
2014-09-11 18:36 ` [PATCH v4 07/16] tools: " Don Slutz
` (9 subsequent siblings)
15 siblings, 1 reply; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
Reduce the VMPORT_DBG_LOG calls.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
xen/arch/x86/hvm/svm/svm.c | 14 ++++++++++++
xen/arch/x86/hvm/vmware/vmport.c | 49 +++++++++++++++++++++++-----------------
xen/arch/x86/hvm/vmx/vmx.c | 12 ++++++++++
xen/include/asm-x86/hvm/trace.h | 45 ++++++++++++++++++++++++++++++++++++
xen/include/asm-x86/hvm/vmport.h | 6 -----
xen/include/public/trace.h | 12 ++++++++++
6 files changed, 111 insertions(+), 27 deletions(-)
diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
index 9e14d2a..2e1ccb9 100644
--- a/xen/arch/x86/hvm/svm/svm.c
+++ b/xen/arch/x86/hvm/svm/svm.c
@@ -2080,6 +2080,14 @@ static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
inst_len = __get_instruction_length_from_list(
v, list, ARRAY_SIZE(list), 0);
+ if ( hvm_long_mode_enabled(v) )
+ HVMTRACE_LONG2_C4D(TRAP_GP, inst_len, regs->rdx,
+ TRC_PAR_LONG(vmcb->exitinfo1),
+ TRC_PAR_LONG(vmcb->exitinfo2));
+ else
+ HVMTRACE_C4D(TRAP_GP, inst_len, vmcb->exitinfo1, vmcb->exitinfo1,
+ vmcb->exitinfo2);
+
rc = vmport_gp_check(regs, v, inst_len, inst_addr,
vmcb->exitinfo1, vmcb->exitinfo2);
if ( !rc )
@@ -2094,6 +2102,12 @@ static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
(unsigned long)vmcb->exitinfo2, regs->rip, inst_addr,
inst_len, regs->rax, regs->rbx, regs->rcx, regs->rdx,
regs->rsi, regs->rdi);
+ if ( hvm_long_mode_enabled(v) )
+ HVMTRACE_LONG_C5D(TRAP_GP_UNKNOWN, rc, regs->rax, regs->rbx, regs->rcx,
+ TRC_PAR_LONG(inst_addr));
+ else
+ HVMTRACE_C5D(TRAP_GP_UNKNOWN, rc, regs->rax, regs->rbx, regs->rcx,
+ inst_addr);
hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
}
}
diff --git a/xen/arch/x86/hvm/vmware/vmport.c b/xen/arch/x86/hvm/vmware/vmport.c
index bfc01dc..26aeb37 100644
--- a/xen/arch/x86/hvm/vmware/vmport.c
+++ b/xen/arch/x86/hvm/vmware/vmport.c
@@ -18,6 +18,7 @@
#include <asm/hvm/hvm.h>
#include <asm/hvm/support.h>
#include <asm/hvm/vmport.h>
+#include <asm/hvm/trace.h>
#include "backdoor_def.h"
#include "guest_msg_def.h"
@@ -69,12 +70,15 @@ int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val)
uint64_t saved_rax = regs->rax;
uint64_t value;
- VMPORT_DBG_LOG(VMPORT_LOG_TRACE,
- "VMware trace dir=%d bytes=%u ip=%"PRIx64" cmd=%d ax=%"
- PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64" si=%"
- PRIx64" di=%"PRIx64"\n", dir, bytes,
- regs->rip, cmd, regs->rax, regs->rbx, regs->rcx,
- regs->rdx, regs->rsi, regs->rdi);
+ if ( dir == IOREQ_READ )
+ HVMTRACE_ND(VMPORT_READ_BEFORE, 0, 1/*cycles*/, 6,
+ regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
+ else
+ HVMTRACE_ND(VMPORT_WRITE_AFTER_BEFORE, 0, 1/*cycles*/, 6,
+ regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
+
switch ( cmd )
{
case BDOOR_CMD_GETMHZ:
@@ -144,19 +148,17 @@ int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val)
regs->rax = 0x0;
break;
default:
- VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
- "VMware bytes=%d dir=%d cmd=%d",
- bytes, dir, cmd);
+ HVMTRACE_ND(VMPORT_UNKNOWN, 0, 1/*cycles*/, 6,
+ (bytes << 8) + dir, cmd, regs->rbx,
+ regs->rcx, regs->rsi, regs->rdi);
break;
}
- VMPORT_DBG_LOG(VMPORT_LOG_VMWARE_AFTER,
- "VMware after ip=%"PRIx64" cmd=%d ax=%"PRIx64" bx=%"
- PRIx64" cx=%"PRIx64" dx=%"PRIx64" si=%"PRIx64" di=%"
- PRIx64"\n",
- regs->rip, cmd, regs->rax, regs->rbx, regs->rcx,
- regs->rdx, regs->rsi, regs->rdi);
+
if ( dir == IOREQ_READ )
{
+ HVMTRACE_ND(VMPORT_READ_AFTER, 0, 1/*cycles*/, 6,
+ regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
switch ( bytes )
{
case 1:
@@ -172,17 +174,22 @@ int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val)
*val = regs->rax;
}
else
+ {
+ HVMTRACE_ND(VMPORT_WRITE_AFTER, 0, 1/*cycles*/, 6,
+ regs->rax, regs->rbx, regs->rcx,
+ regs->rdx, regs->rsi, regs->rdi);
regs->rax = saved_rax;
+ }
}
else
{
+ if ( hvm_long_mode_enabled(current) )
+ HVMTRACE_LONG_C4D(VMPORT_BAD, dir, bytes, regs->rax,
+ TRC_PAR_LONG(regs->rip));
+ else
+ HVMTRACE_C4D(VMPORT_BAD, dir, bytes, regs->rax, regs->rip);
+
rc = X86EMUL_UNHANDLEABLE;
- VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
- "Not VMware %x vs %x; ip=%"PRIx64" ax=%"PRIx64
- " bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64" si=%"PRIx64
- " di=%"PRIx64"",
- magic, BDOOR_MAGIC, regs->rip, regs->rax, regs->rbx,
- regs->rcx, regs->rdx, regs->rsi, regs->rdi);
}
return rc;
diff --git a/xen/arch/x86/hvm/vmx/vmx.c b/xen/arch/x86/hvm/vmx/vmx.c
index 6cd1730..2c678fa 100644
--- a/xen/arch/x86/hvm/vmx/vmx.c
+++ b/xen/arch/x86/hvm/vmx/vmx.c
@@ -2594,6 +2594,12 @@ static void vmx_vmexit_gp_intercept(struct cpu_user_regs *regs,
__vmread(VM_EXIT_INSTRUCTION_LEN, &inst_len);
__vmread(VM_EXIT_INTR_ERROR_CODE, &ecode);
+ if ( hvm_long_mode_enabled(v) )
+ HVMTRACE_LONG2_C4D(TRAP_GP, inst_len, regs->rdx, TRC_PAR_LONG(ecode),
+ TRC_PAR_LONG(exit_qualification));
+ else
+ HVMTRACE_C4D(TRAP_GP, inst_len, regs->rdx, ecode, exit_qualification);
+
rc = vmport_gp_check(regs, v, inst_len, inst_addr,
ecode, exit_qualification);
if ( !rc )
@@ -2607,6 +2613,12 @@ static void vmx_vmexit_gp_intercept(struct cpu_user_regs *regs,
exit_qualification, regs->rip, inst_addr, inst_len,
regs->rax, regs->rbx, regs->rcx, regs->rdx, regs->rsi,
regs->rdi);
+ if ( hvm_long_mode_enabled(v) )
+ HVMTRACE_LONG_C5D(TRAP_GP_UNKNOWN, rc, regs->rax, regs->rbx, regs->rcx,
+ TRC_PAR_LONG(inst_addr));
+ else
+ HVMTRACE_C5D(TRAP_GP_UNKNOWN, rc, regs->rax, regs->rbx, regs->rcx,
+ inst_addr);
hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
}
}
diff --git a/xen/include/asm-x86/hvm/trace.h b/xen/include/asm-x86/hvm/trace.h
index de802a6..8af2d6a 100644
--- a/xen/include/asm-x86/hvm/trace.h
+++ b/xen/include/asm-x86/hvm/trace.h
@@ -52,8 +52,20 @@
#define DO_TRC_HVM_LMSW64 DEFAULT_HVM_MISC
#define DO_TRC_HVM_REALMODE_EMULATE DEFAULT_HVM_MISC
#define DO_TRC_HVM_TRAP DEFAULT_HVM_MISC
+#define DO_TRC_HVM_TRAP64 DEFAULT_HVM_MISC
#define DO_TRC_HVM_TRAP_DEBUG DEFAULT_HVM_MISC
#define DO_TRC_HVM_VLAPIC DEFAULT_HVM_MISC
+#define DO_TRC_HVM_TRAP_GP DEFAULT_HVM_MISC
+#define DO_TRC_HVM_TRAP_GP64 DEFAULT_HVM_MISC
+#define DO_TRC_HVM_TRAP_GP_UNKNOWN DEFAULT_HVM_MISC
+#define DO_TRC_HVM_TRAP_GP_UNKNOWN64 DEFAULT_HVM_MISC
+#define DO_TRC_HVM_VMPORT_READ_BEFORE DEFAULT_HVM_IO
+#define DO_TRC_HVM_VMPORT_WRITE_AFTER_BEFORE DEFAULT_HVM_IO
+#define DO_TRC_HVM_VMPORT_READ_AFTER DEFAULT_HVM_IO
+#define DO_TRC_HVM_VMPORT_WRITE_AFTER DEFAULT_HVM_IO
+#define DO_TRC_HVM_VMPORT_BAD DEFAULT_HVM_IO
+#define DO_TRC_HVM_VMPORT_BAD64 DEFAULT_HVM_IO
+#define DO_TRC_HVM_VMPORT_UNKNOWN DEFAULT_HVM_IO
#define TRC_PAR_LONG(par) ((par)&0xFFFFFFFF),((par)>>32)
@@ -98,6 +110,21 @@
#define HVMTRACE_0D(evt) \
HVMTRACE_ND(evt, 0, 0, 0, 0, 0, 0, 0, 0, 0)
+#define HVMTRACE_C6D(evt, d1, d2, d3, d4, d5, d6) \
+ HVMTRACE_ND(evt, 0, 1, 6, d1, d2, d3, d4, d5, d6)
+#define HVMTRACE_C5D(evt, d1, d2, d3, d4, d5) \
+ HVMTRACE_ND(evt, 0, 1, 5, d1, d2, d3, d4, d5, 0)
+#define HVMTRACE_C4D(evt, d1, d2, d3, d4) \
+ HVMTRACE_ND(evt, 0, 1, 4, d1, d2, d3, d4, 0, 0)
+#define HVMTRACE_C3D(evt, d1, d2, d3) \
+ HVMTRACE_ND(evt, 0, 1, 3, d1, d2, d3, 0, 0, 0)
+#define HVMTRACE_C2D(evt, d1, d2) \
+ HVMTRACE_ND(evt, 0, 1, 2, d1, d2, 0, 0, 0, 0)
+#define HVMTRACE_C1D(evt, d1) \
+ HVMTRACE_ND(evt, 0, 1, 1, d1, 0, 0, 0, 0, 0)
+#define HVMTRACE_C0D(evt) \
+ HVMTRACE_ND(evt, 0, 1, 0, 0, 0, 0, 0, 0, 0)
+
#define HVMTRACE_LONG_1D(evt, d1) \
HVMTRACE_2D(evt ## 64, (d1) & 0xFFFFFFFF, (d1) >> 32)
#define HVMTRACE_LONG_2D(evt, d1, d2, ...) \
@@ -107,6 +134,24 @@
#define HVMTRACE_LONG_4D(evt, d1, d2, d3, d4, ...) \
HVMTRACE_5D(evt ## 64, d1, d2, d3, d4)
+#define HVMTRACE_LONG_C1D(evt, d1) \
+ HVMTRACE_C2D(evt ## 64, (d1) & 0xFFFFFFFF, (d1) >> 32)
+#define HVMTRACE_LONG_C2D(evt, d1, d2, ...) \
+ HVMTRACE_C3D(evt ## 64, d1, d2)
+#define HVMTRACE_LONG_C3D(evt, d1, d2, d3, ...) \
+ HVMTRACE_C4D(evt ## 64, d1, d2, d3)
+#define HVMTRACE_LONG_C4D(evt, d1, d2, d3, d4, ...) \
+ HVMTRACE_C5D(evt ## 64, d1, d2, d3, d4)
+#define HVMTRACE_LONG_C5D(evt, d1, d2, d3, d4, d5, ...) \
+ HVMTRACE_C6D(evt ## 64, d1, d2, d3, d4, d5)
+
+#define HVMTRACE_LONG2_C2D(evt, d1, d2, ...) \
+ HVMTRACE_C4D(evt ## 64, d1, d2)
+#define HVMTRACE_LONG2_C3D(evt, d1, d2, d3, ...) \
+ HVMTRACE_C5D(evt ## 64, d1, d2, d3)
+#define HVMTRACE_LONG2_C4D(evt, d1, d2, d3, d4, ...) \
+ HVMTRACE_C6D(evt ## 64, d1, d2, d3, d4)
+
#endif /* __ASM_X86_HVM_TRACE_H__ */
/*
diff --git a/xen/include/asm-x86/hvm/vmport.h b/xen/include/asm-x86/hvm/vmport.h
index ea2c5a9..74a3b66 100644
--- a/xen/include/asm-x86/hvm/vmport.h
+++ b/xen/include/asm-x86/hvm/vmport.h
@@ -25,12 +25,6 @@
#define VMPORT_LOG_VGP_UNKNOWN (1 << 3)
#define VMPORT_LOG_REALMODE_GP (1 << 4)
-#define VMPORT_LOG_GP_NOT_VMWARE (1 << 9)
-
-#define VMPORT_LOG_TRACE (1 << 16)
-#define VMPORT_LOG_ERROR (1 << 17)
-#define VMPORT_LOG_VMWARE_AFTER (1 << 18)
-
extern unsigned int opt_vmport_debug;
#define VMPORT_DBG_LOG(level, _f, _a...) \
do { \
diff --git a/xen/include/public/trace.h b/xen/include/public/trace.h
index cfcf4aa..ae3613c 100644
--- a/xen/include/public/trace.h
+++ b/xen/include/public/trace.h
@@ -224,11 +224,23 @@
#define TRC_HVM_NPF (TRC_HVM_HANDLER + 0x21)
#define TRC_HVM_REALMODE_EMULATE (TRC_HVM_HANDLER + 0x22)
#define TRC_HVM_TRAP (TRC_HVM_HANDLER + 0x23)
+#define TRC_HVM_TRAP64 (TRC_HVM_HANDLER + TRC_64_FLAG + 0x23)
#define TRC_HVM_TRAP_DEBUG (TRC_HVM_HANDLER + 0x24)
#define TRC_HVM_VLAPIC (TRC_HVM_HANDLER + 0x25)
+#define TRC_HVM_TRAP_GP (TRC_HVM_HANDLER + 0x26)
+#define TRC_HVM_TRAP_GP64 (TRC_HVM_HANDLER + TRC_64_FLAG + 0x26)
+#define TRC_HVM_TRAP_GP_UNKNOWN (TRC_HVM_HANDLER + 0x27)
+#define TRC_HVM_TRAP_GP_UNKNOWN64 (TRC_HVM_HANDLER + TRC_64_FLAG + 0x27)
+#define TRC_HVM_VMPORT_READ_BEFORE (TRC_HVM_HANDLER + 0x28)
+#define TRC_HVM_VMPORT_READ_AFTER (TRC_HVM_HANDLER + 0x29)
+#define TRC_HVM_VMPORT_BAD (TRC_HVM_HANDLER + 0x2a)
+#define TRC_HVM_VMPORT_BAD64 (TRC_HVM_HANDLER + TRC_64_FLAG + 0x2a)
+#define TRC_HVM_VMPORT_UNKNOWN (TRC_HVM_HANDLER + 0x2b)
#define TRC_HVM_IOPORT_WRITE (TRC_HVM_HANDLER + 0x216)
#define TRC_HVM_IOMEM_WRITE (TRC_HVM_HANDLER + 0x217)
+#define TRC_HVM_VMPORT_WRITE_AFTER_BEFORE (TRC_HVM_HANDLER + 0x228)
+#define TRC_HVM_VMPORT_WRITE_AFTER (TRC_HVM_HANDLER + 0x229)
/* Trace events for emulated devices */
#define TRC_HVM_EMUL_HPET_START_TIMER (TRC_HVM_EMUL + 0x1)
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 07/16] tools: Convert vmware_port to xentrace usage
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (5 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 06/16] xen: Convert vmware_port to xentrace usage Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-12 13:15 ` Boris Ostrovsky
2014-09-11 18:36 ` [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc Don Slutz
` (8 subsequent siblings)
15 siblings, 1 reply; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
Also added missing TRAP_DEBUG & VLAPIC.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
tools/xentrace/formats | 13 +++++++++++++
1 file changed, 13 insertions(+)
diff --git a/tools/xentrace/formats b/tools/xentrace/formats
index da658bf..b403c54 100644
--- a/tools/xentrace/formats
+++ b/tools/xentrace/formats
@@ -79,6 +79,19 @@
0x00082020 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) INTR_WINDOW [ value = 0x%(1)08x ]
0x00082021 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) NPF [ gpa = 0x%(2)08x%(1)08x mfn = 0x%(4)08x%(3)08x qual = 0x%(5)04x p2mt = 0x%(6)04x ]
0x00082023 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP [ vector = 0x%(1)02x ]
+0x00082024 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_DEBUG [ exit_qualification = 0x%(1)08x ]
+0x00082025 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VLAPIC
+0x00082026 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP [ inst_len = %(1)d edx = 0x%(2)08x exitinfo1 = 0x%(3)08x exitinfo2 = 0x%(4)08x ]
+0x00082126 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP [ inst_len = %(1)d edx = 0x%(2)08x exitinfo1 = 0x%(4)08x%(3)08x exitinfo2 = 0x%(6)08x%(5)08x ]
+0x00082027 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP_UNKNOWN [ rc = %(1)d eax = 0x%(2)08x ebx = 0x%(3)08x ecx = 0x%(4)08x inst_addr = 0x%(5)08x ]
+0x00082127 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP_UNKNOWN [ rc = %(1)d eax = 0x%(2)08x ebx = 0x%(3)08x ecx = 0x%(4)08x inst_addr = 0x%(6)08x%(5)08x ]
+0x00082028 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_READ_BEFORE [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
+0x00082228 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_WRITE_BEFORE [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
+0x00082029 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_READ_AFTER [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
+0x00082229 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_WRITE_AFTER [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
+0x0008202a CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_BAD [ dir = %(1)d bytes = 0x%(2)d eax = 0x%(3)08x eip = 0x%(4)08x ]
+0x0008212a CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_BAD [ dir = %(1)d bytes = %(2)d eax = 0x%(3)08x rip = 0x%(5)08x%(4)08x ]
+0x0008202b CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_UNKNOWN [ bytes << 8 + dir = 0x%(1)03x cmd = 0x%(2)x cmd = %(2)d ebx = 0x%(3)08x ecx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
0x0010f001 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) page_grant_map [ domid = %(1)d ]
0x0010f002 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) page_grant_unmap [ domid = %(1)d ]
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (6 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 07/16] tools: " Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-12 13:37 ` Boris Ostrovsky
2014-09-11 18:36 ` [PATCH v4 09/16] tools: " Don Slutz
` (7 subsequent siblings)
15 siblings, 1 reply; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
The support included is enough to allow VMware tools to install in a
HVM domU and provide guestinfo support. guestinfo support is
provide by what is known as VMware RPC support. This guestinfo
support is provided via libxc. libxl support has not be written.
Note: VMware RPC support is only available on HVM domU.
This interface is an extension of __HYPERVISOR_HVM_op. It was
picked because xc_get_hvm_param() also uses it and VMware guest
info is a lot like a hvm param.
The HVMOP_get_vmport_guest_info is used by two libxc functions,
xc_get_vmport_guest_info and xc_fetch_all_vmport_guest_info.
xc_fetch_all_vmport_guest_info is designed to be used to fetch all
currently set guestinfo values.
To save on hypervisor heap memory, the guestinfo support in done in
two sizes, normal and jumbo. Normal is used to handle up to 128
byte values and jumbo is used to handle up to 4096 byte values.
Since all this is work is done when the guest is doing a single
instruction; it was designed to not use the hypervisor heap to
allocate the memory at this time. Instead a few are allocated at
the create domain time and during the xen's hyper-call to get or set
them. This was picked in that if a tool stack is using the VMware
guest info support, it should be using either of both of the get and
set. And so in this case the guest should only see an out of memory
error when the compile max amount of hypervisor heap memory is in
use.
Doing it this way does lead to a lot of pointer use and many
sub structures.
If the domU is running VMware tools, then the "build version" of
the tools is also available via xc_get_HVM_param(). This also
enables the use of new triggers that will use the VMware hyper-call
to do some limited control of the domU. The most useful are
poweroff and reboot. Since a guest process needs to be running
for these to work, a tool stack should check that the build version
is non zero before assuming these will work.
The 2 hvm param's HVM_PARAM_VMPORT_BUILD_NUMBER_TIME and
HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE are how "build version" is
accessed. These 2 params are only allowed to be set to zero. The
HVM_PARAM_VMPORT_BUILD_NUMBER_TIME can be used to track the last
time the VMware tools in the guest responded. One such use would
be the health of the tools in the guest. The hvm param
HVM_PARAM_VMPORT_RESET_TIME controls how often to request them in
seconds minus 1. The minus 1 is to handle to 0 case. I.E. the
fastest that can be selected is every second. The default is 4
times a minute.
The VMware RPC support includes the notion of channels that are
opened, active and closed. All RPC messages sent via a channel
starts with normal ASCII text. The message some times does include
binary data.
Currently there are 2 protocols defined for VMware RPC. They
determine the direction for data flow, domU to tool stack or
tool stack to domU.
There is no provided interrupt for VMware RPC.
For a debug=y build there is a new command line option
vmport_debug=. It enabled output to the console of various
stages of handling the "IN EAX, DX" instruction. Most uses
are the summary ones that show complete RPC actions.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
xen/arch/x86/hvm/hvm.c | 43 +
xen/arch/x86/hvm/vmware/Makefile | 1 +
xen/arch/x86/hvm/vmware/vmport.c | 7 +
xen/arch/x86/hvm/vmware/vmport_rpc.c | 1281 ++++++++++++++++++++++++++
xen/include/asm-x86/hvm/domain.h | 4 +
xen/include/asm-x86/hvm/vmport.h | 16 +
xen/include/public/arch-x86/hvm/vmporttype.h | 118 +++
xen/include/public/hvm/hvm_op.h | 18 +
xen/include/public/hvm/params.h | 5 +-
9 files changed, 1492 insertions(+), 1 deletion(-)
create mode 100644 xen/arch/x86/hvm/vmware/vmport_rpc.c
create mode 100644 xen/include/public/arch-x86/hvm/vmporttype.h
diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
index 9c59c85..536bec4 100644
--- a/xen/arch/x86/hvm/hvm.c
+++ b/xen/arch/x86/hvm/hvm.c
@@ -1500,7 +1500,12 @@ int hvm_domain_initialise(struct domain *d)
d->arch.hvm_domain.io_handler->num_slot = 0;
if ( d->arch.hvm_domain.is_vmware_port_enabled )
+ {
vmport_register(d);
+ rc = vmport_rpc_init(d);
+ if ( rc != 0 )
+ goto fail1;
+ }
if ( is_pvh_domain(d) )
{
@@ -1537,6 +1542,7 @@ int hvm_domain_initialise(struct domain *d)
stdvga_deinit(d);
vioapic_deinit(d);
fail1:
+ vmport_rpc_deinit(d);
xfree(d->arch.hvm_domain.io_handler);
xfree(d->arch.hvm_domain.params);
fail0:
@@ -6142,6 +6148,43 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
break;
}
+ case HVMOP_get_vmport_guest_info:
+ case HVMOP_set_vmport_guest_info:
+ {
+ struct xen_hvm_vmport_guest_info a;
+ struct domain *d;
+
+ if ( copy_from_guest(&a, arg, 1) )
+ return -EFAULT;
+
+ rc = vmport_rpc_hvmop_precheck(op, &a);
+ if ( rc )
+ return rc;
+
+ d = rcu_lock_domain_by_any_id(a.domid);
+ if ( d == NULL )
+ return rc;
+
+ rc = -EINVAL;
+ if ( !is_hvm_domain(d) )
+ goto param_fail9;
+
+ rc = xsm_hvm_param(XSM_TARGET, d, op);
+ if ( rc )
+ goto param_fail9;
+
+ rc = vmport_rpc_hvmop_do(d, op, &a);
+ if ( rc )
+ goto param_fail9;
+
+ if ( op == HVMOP_get_vmport_guest_info )
+ rc = copy_to_guest(arg, &a, 1) ? -EFAULT : 0;
+
+ param_fail9:
+ rcu_unlock_domain(d);
+ break;
+ }
+
default:
{
gdprintk(XENLOG_DEBUG, "Bad HVM op %ld.\n", op);
diff --git a/xen/arch/x86/hvm/vmware/Makefile b/xen/arch/x86/hvm/vmware/Makefile
index cd8815b..4a14124 100644
--- a/xen/arch/x86/hvm/vmware/Makefile
+++ b/xen/arch/x86/hvm/vmware/Makefile
@@ -1,2 +1,3 @@
obj-y += cpuid.o
obj-y += vmport.o
+obj-y += vmport_rpc.o
diff --git a/xen/arch/x86/hvm/vmware/vmport.c b/xen/arch/x86/hvm/vmware/vmport.c
index 26aeb37..9e308a0 100644
--- a/xen/arch/x86/hvm/vmware/vmport.c
+++ b/xen/arch/x86/hvm/vmware/vmport.c
@@ -139,6 +139,13 @@ int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val)
/* maxTimeLag */
regs->rcx = 0;
break;
+ case BDOOR_CMD_MESSAGE:
+ if ( !is_pv_vcpu(current) )
+ {
+ /* Only supported for non pv domains */
+ vmport_rpc(¤t->domain->arch.hvm_domain, regs);
+ }
+ break;
case BDOOR_CMD_GETGUIOPTIONS:
regs->rax = VMWARE_GUI_AUTO_GRAB | VMWARE_GUI_AUTO_UNGRAB |
VMWARE_GUI_AUTO_RAISE_DISABLED | VMWARE_GUI_SYNC_TIME |
diff --git a/xen/arch/x86/hvm/vmware/vmport_rpc.c b/xen/arch/x86/hvm/vmware/vmport_rpc.c
new file mode 100644
index 0000000..695e58f
--- /dev/null
+++ b/xen/arch/x86/hvm/vmware/vmport_rpc.c
@@ -0,0 +1,1281 @@
+/*
+ * HVM VMPORT RPC emulation
+ *
+ * Copyright (C) 2012 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+/*
+ * VMware Tools running in a DOMU will do "info-get" and "info-set"
+ * guestinfo commands to get and set keys and values. Inside the VM,
+ * vmtools at its lower level will feed the command string 4 bytes
+ * at a time into the VMWARE magic port using the IN
+ * instruction. Each 4 byte mini-rpc will get handled
+ * vmport_io()-->vmport_rpc()-->vmport_process_packet()-->
+ * vmport_process_send_payload()-->vmport_send() and the command
+ * string will get accumulated into a channels send_buffer. When
+ * the full length of the string has been accumulated, then this
+ * code copies the send_buffer into a free
+ * vmport_state->channel-->receive_bucket.buffer
+ * VMware tools then does RECVSIZE and RECVPAYLOAD messages, the
+ * latter then reads 4 bytes at a time using the IN instruction (for
+ * the info-get case). Then a final RECVSTATUS message is sent to
+ * finish up
+ */
+
+#include <xen/config.h>
+#include <xen/lib.h>
+#include <xen/cper.h>
+#include <asm/hvm/hvm.h>
+#include <asm/hvm/support.h>
+#include <asm/hvm/vmport.h>
+#include <asm/hvm/trace.h>
+
+#include "public/arch-x86/hvm/vmporttype.h"
+
+#include "backdoor_def.h"
+#include "guest_msg_def.h"
+
+
+#define VMWARE_PROTO_TO_GUEST 0x4f4c4354
+#define VMWARE_PROTO_FROM_GUEST 0x49435052
+
+#define GUESTINFO_NOTFOUND 500
+#define GUESTINFO_VALTOOLONG 1
+#define GUESTINFO_KEYTOOLONG 2
+#define GUESTINFO_TOOMANYKEYS 3
+
+
+static inline uint16_t get_low_bits(uint32_t bits)
+{
+ return bits & 0xffff;
+}
+
+static inline uint16_t get_high_bits(uint32_t bits)
+{
+ return bits >> 16;
+}
+
+static inline uint32_t set_high_bits(uint32_t b, uint32_t val)
+{
+ return (val << 16) | get_low_bits(b);
+}
+
+static inline void set_status(struct cpu_user_regs *ur, uint16_t val)
+{
+ /* VMware defines this to be only 32 bits */
+ ur->rcx = (val << 16) | (ur->rcx & 0xffff);
+}
+
+#ifndef NDEBUG
+static void vmport_safe_print(char *prefix, int len, const char *msg)
+{
+ unsigned char c;
+ unsigned int end = len;
+ unsigned int i, k;
+ char out[4 * (VMPORT_MAX_SEND_BUF + 1) * 3 + 6];
+
+ if ( end > (sizeof(out) / 3 - 6) )
+ end = sizeof(out) / 3 - 6;
+ out[0] = '<';
+ k = 1;
+ for ( i = 0; i < end; i++ )
+ {
+ c = msg[i];
+ if ( (c == '^') || (c == '\\') || (c == '>') )
+ {
+ out[k++] = '\\';
+ out[k++] = c;
+ }
+ else if ( (c >= ' ') && (c <= '~') )
+ out[k++] = c;
+ else if ( c < ' ' )
+ {
+ out[k++] = '^';
+ out[k++] = c ^ 0x40;
+ }
+ else
+ {
+ snprintf(&out[k], sizeof(out) - k, "\\%02x", c);
+ k += 3;
+ }
+ }
+ out[k++] = '>';
+ if ( len > end )
+ {
+ out[k++] = '.';
+ out[k++] = '.';
+ out[k++] = '.';
+ }
+ out[k++] = 0;
+ gdprintk(XENLOG_DEBUG, "%s%d(%d,%d,%zu)%s\n", prefix, end, len, k,
+ sizeof(out), out);
+}
+#endif
+
+/*
+ * Copy message into a jumbo bucket buffer which vmtools will use to
+ * read from 4 bytes at a time until done with it
+ */
+static void vmport_send_jumbo(struct hvm_domain *hd, vmport_channel_t *c,
+ const char *msg)
+{
+ unsigned int cur_recv_len = strlen(msg) + 1;
+ vmport_jumbo_bucket_t *b = &(c->jumbo_recv_bkt);
+
+ b->ctl.recv_len = cur_recv_len;
+ b->ctl.recv_idx = 0;
+
+ memset(b->recv_buf, 0, sizeof(b->recv_buf));
+
+ if ( cur_recv_len >= (sizeof(b->recv_buf) - 1) )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
+ "VMware jumbo recv_len=%d >= %ld",
+ cur_recv_len, sizeof(b->recv_buf) - 1);
+ cur_recv_len = sizeof(b->recv_buf) - 1;
+ }
+
+ memcpy(b->recv_buf, msg, cur_recv_len);
+
+ c->ctl.jumbo = 1;
+}
+
+/*
+ * Copy message into a free receive bucket buffer which vmtools will use to
+ * read from 4 bytes at a time until done with it
+ */
+static void vmport_send_normal(struct hvm_domain *hd, vmport_channel_t *c,
+ const char *msg)
+{
+ unsigned int cur_recv_len = strlen(msg) + 1;
+ unsigned int my_bkt = c->ctl.recv_write;
+ unsigned int next_bkt = my_bkt + 1;
+ vmport_bucket_t *b;
+
+ if ( next_bkt >= VMPORT_MAX_BKTS )
+ next_bkt = 0;
+
+ if ( next_bkt == c->ctl.recv_read )
+ {
+#ifndef NDEBUG
+ if ( opt_vmport_debug & VMPORT_LOG_SKIP_SEND )
+ {
+ char prefix[30];
+
+ snprintf(prefix, sizeof(prefix),
+ "VMware _send skipped %d (%d, %d) ",
+ c->ctl.chan_id, my_bkt, c->ctl.recv_read);
+ prefix[sizeof(prefix) - 1] = 0;
+ vmport_safe_print(prefix, cur_recv_len, msg);
+ }
+#endif
+ return;
+ }
+
+ c->ctl.recv_write = next_bkt;
+ b = &c->recv_bkt[my_bkt];
+#ifndef NDEBUG
+ if ( opt_vmport_debug & VMPORT_LOG_SEND )
+ {
+ char prefix[30];
+
+ snprintf(prefix, sizeof(prefix), "VMware _send %d (%d) ",
+ c->ctl.chan_id, my_bkt);
+ prefix[sizeof(prefix) - 1] = 0;
+ vmport_safe_print(prefix, cur_recv_len, msg);
+ }
+#endif
+
+ b->ctl.recv_len = cur_recv_len;
+ b->ctl.recv_idx = 0;
+ memset(b->recv_buf, 0, sizeof(b->recv_buf));
+ if ( cur_recv_len >= (sizeof(b->recv_buf) - 1) )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware recv_len=%d >= %zd",
+ cur_recv_len, sizeof(b->recv_buf) - 1);
+ cur_recv_len = sizeof(b->recv_buf) - 1;
+ }
+ memcpy(b->recv_buf, msg, cur_recv_len);
+}
+
+static void vmport_send(struct hvm_domain *hd, vmport_channel_t *c,
+ const char *msg)
+{
+ unsigned int cur_recv_len = strlen(msg) + 1;
+
+ if ( cur_recv_len > VMPORT_MAX_VAL_LEN )
+ vmport_send_jumbo(hd, c, msg);
+ else
+ vmport_send_normal(hd, c, msg);
+}
+
+void vmport_ctrl_send(struct hvm_domain *hd, char *msg)
+{
+ struct vmport_state *vs = hd->vmport_data;
+ unsigned int i;
+
+ hd->vmport_data->ping_time = get_sec();
+ spin_lock(&hd->vmport_lock);
+ for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
+ {
+ if ( vs->chans[i].ctl.proto_num == VMWARE_PROTO_TO_GUEST )
+ vmport_send(hd, &vs->chans[i], msg);
+ }
+ spin_unlock(&hd->vmport_lock);
+}
+
+static void vmport_flush(struct hvm_domain *hd)
+{
+ spin_lock(&hd->vmport_lock);
+ memset(&hd->vmport_data->chans, 0, sizeof(hd->vmport_data->chans));
+ spin_unlock(&hd->vmport_lock);
+}
+
+static void vmport_sweep(struct hvm_domain *hd, unsigned long now_time)
+{
+ struct vmport_state *vs = hd->vmport_data;
+ unsigned int i;
+
+ for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
+ {
+ if ( vs->chans[i].ctl.proto_num )
+ {
+ vmport_channel_t *c = &vs->chans[i];
+ long delta = now_time - c->ctl.active_time;
+
+ if ( delta >= 80 )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_SWEEP, "VMware flush %d. delta=%ld",
+ c->ctl.chan_id, delta);
+ /* Return channel to free pool */
+ c->ctl.proto_num = 0;
+ }
+ }
+ }
+}
+
+static vmport_channel_t *vmport_new_chan(struct vmport_state *vs,
+ unsigned long now_time)
+{
+ unsigned int i;
+
+ for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
+ {
+ if ( !vs->chans[i].ctl.proto_num )
+ {
+ vmport_channel_t *c = &vs->chans[i];
+
+ c->ctl.chan_id = i;
+ c->ctl.cookie = vs->open_cookie++;
+ c->ctl.active_time = now_time;
+ c->ctl.send_len = 0;
+ c->ctl.send_idx = 0;
+ c->ctl.recv_read = 0;
+ c->ctl.recv_write = 0;
+ return c;
+ }
+ }
+ return NULL;
+}
+
+static void vmport_process_send_size(struct hvm_domain *hd, vmport_channel_t *c,
+ struct cpu_user_regs *ur)
+{
+ /* vmware tools often send a 0 byte request size. */
+ c->ctl.send_len = ur->rbx;
+ c->ctl.send_idx = 0;
+
+ set_status(ur, MESSAGE_STATUS_SUCCESS);
+}
+
+/* ret_buffer is in/out param */
+static int vmport_get_guestinfo(struct hvm_domain *hd, struct vmport_state *vs,
+ char *a_info_key, unsigned int a_key_len,
+ char *ret_buffer, unsigned int ret_buffer_len)
+{
+ unsigned int i;
+
+ for ( i = 0; i < vs->used_guestinfo; i++ )
+ {
+ if ( vs->guestinfo[i] &&
+ (vs->guestinfo[i]->key_len == a_key_len) &&
+ (memcmp(a_info_key, vs->guestinfo[i]->key_data,
+ vs->guestinfo[i]->key_len) == 0) )
+ {
+ snprintf(ret_buffer, ret_buffer_len - 1, "1 %.*s",
+ (int)vs->guestinfo[i]->val_len,
+ vs->guestinfo[i]->val_data);
+ return i;
+ }
+ }
+
+ for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
+ {
+ if ( vs->guestinfo_jumbo[i] &&
+ (vs->guestinfo_jumbo[i]->key_len == a_key_len) &&
+ (memcmp(a_info_key, vs->guestinfo_jumbo[i]->key_data,
+ vs->guestinfo_jumbo[i]->key_len) == 0) )
+ {
+ snprintf(ret_buffer, ret_buffer_len - 1, "1 %.*s",
+ (int)vs->guestinfo_jumbo[i]->val_len,
+ vs->guestinfo_jumbo[i]->val_data);
+ return i;
+ }
+ }
+ return GUESTINFO_NOTFOUND;
+}
+
+static void hvm_del_guestinfo_jumbo(struct vmport_state *vs, char *key,
+ uint8_t len)
+{
+ int i;
+
+ for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
+ {
+ if ( !vs->guestinfo_jumbo[i] )
+ {
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING,
+ "i=%d not allocated used_guestinfo_jumbo=%d\n",
+ i, vs->used_guestinfo_jumbo);
+#endif
+ }
+ else if ( (vs->guestinfo_jumbo[i]->key_len == len) &&
+ (memcmp(key, vs->guestinfo_jumbo[i]->key_data, len) == 0) )
+ {
+ vs->guestinfo_jumbo[i]->key_len = 0;
+ vs->guestinfo_jumbo[i]->val_len = 0;
+ break;
+ }
+ }
+}
+
+static void hvm_del_guestinfo(struct vmport_state *vs, char *key, uint8_t len)
+{
+ int i;
+
+ for ( i = 0; i < vs->used_guestinfo; i++ )
+ {
+ if ( !vs->guestinfo[i] )
+ {
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING,
+ "i=%d not allocated, but used_guestinfo=%d\n",
+ i, vs->used_guestinfo);
+#endif
+ }
+ else if ( (vs->guestinfo[i]->key_len == len) &&
+ (memcmp(key, vs->guestinfo[i]->key_data, len) == 0) )
+ {
+ vs->guestinfo[i]->key_len = 0;
+ vs->guestinfo[i]->val_len = 0;
+ break;
+ }
+ }
+}
+
+static int vmport_set_guestinfo(struct vmport_state *vs, int a_key_len,
+ unsigned int a_val_len, char *a_info_key, char *val)
+{
+ unsigned int i;
+ int free_i = -1, rc = 0;
+
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING, "vmport_set_guestinfo a_val_len=%d\n", a_val_len);
+#endif
+
+ if ( a_key_len <= VMPORT_MAX_KEY_LEN )
+ {
+ if ( a_val_len <= VMPORT_MAX_VAL_LEN )
+ {
+ for ( i = 0; i < vs->used_guestinfo; i++ )
+ {
+ if ( !vs->guestinfo[i] )
+ {
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING,
+ "i=%d not allocated, but used_guestinfo=%d\n",
+ i, vs->used_guestinfo);
+#endif
+ }
+ else if ( (vs->guestinfo[i]->key_len == a_key_len) &&
+ (memcmp(a_info_key, vs->guestinfo[i]->key_data,
+ vs->guestinfo[i]->key_len) == 0) )
+ {
+ vs->guestinfo[i]->val_len = a_val_len;
+ memcpy(vs->guestinfo[i]->val_data, val, a_val_len);
+ break;
+ }
+ else if ( (vs->guestinfo[i]->key_len == 0) &&
+ (free_i == -1) )
+ free_i = i;
+ }
+ if ( i >= vs->used_guestinfo )
+ {
+ if ( free_i == -1 )
+ rc = GUESTINFO_TOOMANYKEYS;
+ else
+ {
+ vs->guestinfo[free_i]->key_len = a_key_len;
+ memcpy(vs->guestinfo[free_i]->key_data,
+ a_info_key, a_key_len);
+ vs->guestinfo[free_i]->val_len = a_val_len;
+ memcpy(vs->guestinfo[free_i]->val_data,
+ val, a_val_len);
+ }
+ }
+ }
+ else
+ rc = GUESTINFO_VALTOOLONG;
+ }
+ else
+ rc = GUESTINFO_KEYTOOLONG;
+ if ( !rc )
+ hvm_del_guestinfo_jumbo(vs, a_info_key, a_key_len);
+ return rc;
+}
+
+static int vmport_set_guestinfo_jumbo(struct vmport_state *vs, int a_key_len,
+ int a_val_len, char *a_info_key, char *val)
+{
+ unsigned int i;
+ int free_i = -1, rc = 0;
+
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING, "vmport_set_guestinfo_jumbo a_val_len=%d\n",
+ a_val_len);
+#endif
+
+ if ( a_key_len <= VMPORT_MAX_KEY_LEN )
+ {
+ if ( a_val_len <= VMPORT_MAX_VAL_JUMBO_LEN )
+ {
+ for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
+ {
+ if ( !vs->guestinfo_jumbo[i] )
+ {
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING,
+ "i=%d not allocated; used_guestinfo_jumbo=%d\n",
+ i, vs->used_guestinfo_jumbo);
+#endif
+ }
+ else if ( (vs->guestinfo_jumbo[i]->key_len == a_key_len) &&
+ (memcmp(a_info_key,
+ vs->guestinfo_jumbo[i]->key_data,
+ vs->guestinfo_jumbo[i]->key_len) == 0) )
+ {
+
+ vs->guestinfo_jumbo[i]->val_len = a_val_len;
+ memcpy(vs->guestinfo_jumbo[i]->val_data, val, a_val_len);
+ break;
+ }
+ else if ( (vs->guestinfo_jumbo[i]->key_len == 0) &&
+ (free_i == -1) )
+ free_i = i;
+ }
+ if ( i >= vs->used_guestinfo_jumbo )
+ {
+ if ( free_i == -1 )
+ rc = GUESTINFO_TOOMANYKEYS;
+ else
+ {
+ vs->guestinfo_jumbo[free_i]->key_len = a_key_len;
+ memcpy(vs->guestinfo_jumbo[free_i]->key_data,
+ a_info_key, a_key_len);
+ vs->guestinfo_jumbo[free_i]->val_len = a_val_len;
+ memcpy(vs->guestinfo_jumbo[free_i]->val_data,
+ val, a_val_len);
+ }
+ }
+ }
+ else
+ rc = GUESTINFO_VALTOOLONG;
+ }
+ else
+ rc = GUESTINFO_KEYTOOLONG;
+ if ( !rc )
+ hvm_del_guestinfo(vs, a_info_key, a_key_len);
+ return rc;
+}
+
+static void vmport_process_send_payload(struct hvm_domain *hd,
+ vmport_channel_t *c,
+ struct cpu_user_regs *ur,
+ unsigned long now_time)
+{
+ /* Accumulate 4 bytes of paload into send_buf using offset */
+ if ( c->ctl.send_idx < VMPORT_MAX_SEND_BUF )
+ c->send_buf[c->ctl.send_idx] = ur->rbx;
+
+ c->ctl.send_idx++;
+ set_status(ur, MESSAGE_STATUS_SUCCESS);
+
+ if ( c->ctl.send_idx * 4 >= c->ctl.send_len )
+ {
+
+ /* We are done accumulating so handle the command */
+
+ if ( c->ctl.send_idx < VMPORT_MAX_SEND_BUF )
+ ((char *)c->send_buf)[c->ctl.send_len] = 0;
+#ifndef NDEBUG
+ if ( opt_vmport_debug & VMPORT_LOG_RECV )
+ {
+ char prefix[30];
+
+ snprintf(prefix, sizeof(prefix),
+ "VMware RECV %d (%d) ", c->ctl.chan_id, c->ctl.recv_read);
+ prefix[sizeof(prefix) - 1] = 0;
+ vmport_safe_print(prefix, c->ctl.send_len, (char *)c->send_buf);
+ }
+#endif
+ if ( c->ctl.proto_num == VMWARE_PROTO_FROM_GUEST )
+ {
+ /*
+ * Eaxmples of messages:
+ *
+ * log toolbox: Version: build-341836
+ * SetGuestInfo 4 build-341836
+ * info-get guestinfo.ip
+ * info-set guestinfo.ip joe
+ *
+ */
+
+ char *build = NULL;
+ char *info_key = NULL;
+ char *ret_msg = "1 ";
+ char ret_buffer[2 + VMPORT_MAX_VAL_JUMBO_LEN + 2];
+
+ if ( strncmp((char *)c->send_buf, "log toolbox: Version: build-",
+ strlen("log toolbox: Version: build-")) == 0 )
+
+ build = (char *)c->send_buf +
+ strlen("log toolbox: Version: build-");
+
+ else if ( strncmp((char *)c->send_buf, "SetGuestInfo 4 build-",
+ strlen("SetGuestInfo 4 build-")) == 0 )
+
+ build = (char *)c->send_buf + strlen("SetGuestInfo 4 build-");
+
+ else if ( strncmp((char *)c->send_buf, "info-get guestinfo.",
+ strlen("info-get guestinfo.")) == 0 )
+ {
+
+ unsigned int a_key_len = c->ctl.send_len -
+ strlen("info-get guestinfo.");
+ int rc;
+ struct vmport_state *vs = hd->vmport_data;
+
+ info_key = (char *)c->send_buf + strlen("info-get guestinfo.");
+ if ( a_key_len <= VMPORT_MAX_KEY_LEN )
+ {
+
+ rc = vmport_get_guestinfo(hd, vs, info_key, a_key_len,
+ ret_buffer, sizeof(ret_buffer));
+ if ( rc == GUESTINFO_NOTFOUND )
+ ret_msg = "0 No value found";
+ else
+ ret_msg = ret_buffer;
+ }
+ else
+ ret_msg = "0 Key is too long";
+
+ }
+ else if ( strncmp((char *)c->send_buf, "info-set guestinfo.",
+ strlen("info-set guestinfo.")) == 0 )
+ {
+ char *val;
+ unsigned int rest_len = c->ctl.send_len -
+ strlen("info-set guestinfo.");
+
+ info_key = (char *)c->send_buf + strlen("info-set guestinfo.");
+ val = strstr(info_key, " ");
+ if ( val )
+ {
+ unsigned int a_key_len = val - info_key;
+ unsigned int a_val_len = rest_len - a_key_len - 1;
+ int rc;
+ struct vmport_state *vs = hd->vmport_data;
+
+ val++;
+ if ( a_val_len > VMPORT_MAX_VAL_LEN )
+ rc = vmport_set_guestinfo_jumbo(vs, a_key_len,
+ a_val_len,
+ info_key, val);
+ else
+ rc = vmport_set_guestinfo(vs, a_key_len, a_val_len,
+ info_key, val);
+ if ( rc == 0 )
+ ret_msg = "1 ";
+ if ( rc == GUESTINFO_VALTOOLONG )
+ ret_msg = "0 Value too long";
+ if ( rc == GUESTINFO_KEYTOOLONG )
+ ret_msg = "0 Key is too long";
+ if ( rc == GUESTINFO_TOOMANYKEYS )
+ ret_msg = "0 Too many keys";
+
+
+ }
+ else
+ ret_msg = "0 Two and exactly two arguments expected";
+ }
+
+ vmport_send(hd, c, ret_msg);
+ if ( build )
+ {
+ long val = 0;
+ char *p = build;
+
+ while ( *p )
+ {
+ if ( *p < '0' || *p > '9' )
+ break;
+ val = val * 10 + *p - '0';
+ p++;
+ };
+
+ hd->params[HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE] = val;
+ hd->params[HVM_PARAM_VMPORT_BUILD_NUMBER_TIME] = now_time;
+ }
+ }
+ else
+ {
+ unsigned int my_bkt = c->ctl.recv_read - 1;
+ vmport_bucket_t *b;
+
+ if ( my_bkt >= VMPORT_MAX_BKTS )
+ my_bkt = VMPORT_MAX_BKTS - 1;
+ b = &c->recv_bkt[my_bkt];
+ b->ctl.recv_len = 0;
+ }
+ }
+}
+
+static void vmport_process_recv_size(struct hvm_domain *hd, vmport_channel_t *c,
+ struct cpu_user_regs *ur)
+{
+ vmport_bucket_t *b;
+ vmport_jumbo_bucket_t *jb;
+ int16_t recv_len;
+
+ if ( c->ctl.jumbo )
+ {
+ jb = &c->jumbo_recv_bkt;
+ recv_len = jb->ctl.recv_len;
+ }
+ else
+ {
+ b = &c->recv_bkt[c->ctl.recv_read];
+ recv_len = b->ctl.recv_len;
+ }
+ if ( recv_len )
+ {
+ set_status(ur, MESSAGE_STATUS_DORECV | MESSAGE_STATUS_SUCCESS);
+ ur->rdx = set_high_bits(ur->rdx, MESSAGE_TYPE_SENDSIZE);
+ ur->rbx = recv_len;
+ }
+ else
+ set_status(ur, MESSAGE_STATUS_SUCCESS);
+}
+
+static void vmport_process_recv_payload(struct hvm_domain *hd,
+ vmport_channel_t *c,
+ struct cpu_user_regs *ur)
+{
+ vmport_bucket_t *b;
+ vmport_jumbo_bucket_t *jb;
+
+ if ( c->ctl.jumbo )
+ {
+ jb = &c->jumbo_recv_bkt;
+ ur->rbx = jb->recv_buf[jb->ctl.recv_idx++];
+ }
+ else
+ {
+ b = &c->recv_bkt[c->ctl.recv_read];
+ if ( b->ctl.recv_idx < VMPORT_MAX_RECV_BUF )
+ ur->rbx = b->recv_buf[b->ctl.recv_idx++];
+ else
+ ur->rbx = 0;
+ }
+
+ set_status(ur, MESSAGE_STATUS_SUCCESS);
+ ur->rdx = set_high_bits(ur->rdx, MESSAGE_TYPE_SENDPAYLOAD);
+}
+
+static void vmport_process_recv_status(struct hvm_domain *hd,
+ vmport_channel_t *c,
+ struct cpu_user_regs *ur)
+{
+ vmport_bucket_t *b;
+ vmport_jumbo_bucket_t *jb;
+
+ set_status(ur, MESSAGE_STATUS_SUCCESS);
+
+ if ( c->ctl.jumbo )
+ {
+ c->ctl.jumbo = 0;
+ /* add debug here */
+ jb = &c->jumbo_recv_bkt;
+ return;
+ }
+
+ b = &c->recv_bkt[c->ctl.recv_read];
+
+ c->ctl.recv_read++;
+ if ( c->ctl.recv_read >= VMPORT_MAX_BKTS )
+ c->ctl.recv_read = 0;
+}
+
+static void vmport_process_close(struct hvm_domain *hd, vmport_channel_t *c,
+ struct cpu_user_regs *ur)
+{
+ /* Return channel to free pool */
+ c->ctl.proto_num = 0;
+ set_status(ur, MESSAGE_STATUS_SUCCESS);
+}
+
+static void vmport_process_packet(struct hvm_domain *hd, vmport_channel_t *c,
+ struct cpu_user_regs *ur, unsigned int sub_cmd,
+ unsigned long now_time)
+{
+ c->ctl.active_time = now_time;
+
+ switch ( sub_cmd )
+ {
+ case MESSAGE_TYPE_SENDSIZE:
+ vmport_process_send_size(hd, c, ur);
+ break;
+
+ case MESSAGE_TYPE_SENDPAYLOAD:
+ vmport_process_send_payload(hd, c, ur, now_time);
+ break;
+
+ case MESSAGE_TYPE_RECVSIZE:
+ vmport_process_recv_size(hd, c, ur);
+ break;
+
+ case MESSAGE_TYPE_RECVPAYLOAD:
+ vmport_process_recv_payload(hd, c, ur);
+ break;
+
+ case MESSAGE_TYPE_RECVSTATUS:
+ vmport_process_recv_status(hd, c, ur);
+ break;
+
+ case MESSAGE_TYPE_CLOSE:
+ vmport_process_close(hd, c, ur);
+ break;
+
+ default:
+ ur->rcx = 0;
+ break;
+ }
+}
+
+void vmport_rpc(struct hvm_domain *hd, struct cpu_user_regs *ur)
+{
+ unsigned int sub_cmd = get_high_bits(ur->rcx);
+ vmport_channel_t *c = NULL;
+ uint16_t msg_id;
+ uint32_t msg_cookie;
+ unsigned long now_time = get_sec();
+ long delta = now_time - hd->vmport_data->ping_time;
+
+ if ( delta > hd->params[HVM_PARAM_VMPORT_RESET_TIME] )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_PING, "VMware ping. delta=%ld",
+ delta);
+ vmport_ctrl_send(hd, "reset");
+ }
+ spin_lock(&hd->vmport_lock);
+ vmport_sweep(hd, now_time);
+ do {
+ /* Check to see if a new open request is happening... */
+ if ( MESSAGE_TYPE_OPEN == sub_cmd )
+ {
+ c = vmport_new_chan(hd->vmport_data, now_time);
+ if ( NULL == c )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
+ "VMware failed to find a free channel");
+ break;
+ }
+
+ /* Attach the apropriate protocol the the channel */
+ c->ctl.proto_num = ur->rbx & ~GUESTMSG_FLAG_COOKIE;
+ set_status(ur, MESSAGE_STATUS_SUCCESS);
+ ur->rdx = set_high_bits(ur->rdx, c->ctl.chan_id);
+ ur->rdi = get_low_bits(c->ctl.cookie);
+ ur->rsi = get_high_bits(c->ctl.cookie);
+ if ( c->ctl.proto_num == VMWARE_PROTO_TO_GUEST )
+ vmport_send(hd, c, "reset");
+ break;
+ }
+
+ msg_id = get_high_bits(ur->rdx);
+ msg_cookie = set_high_bits(ur->rdi, ur->rsi);
+ if ( msg_id >= VMPORT_MAX_CHANS )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware chan id err %d >= %d",
+ msg_id, VMPORT_MAX_CHANS);
+ break;
+ }
+ c = &hd->vmport_data->chans[msg_id];
+ if ( !c->ctl.proto_num )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware chan %d not open",
+ msg_id);
+ break;
+ }
+
+ /* We check the cookie here since it's possible that the
+ * connection timed out on us and another channel was opened
+ * if this happens, return error and the um tool will
+ * need to reopen the connection
+ */
+ if ( msg_cookie != c->ctl.cookie )
+ {
+ VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware ctl.cookie err %x vs %x",
+ msg_cookie, c->ctl.cookie);
+ break;
+ }
+ vmport_process_packet(hd, c, ur, sub_cmd, now_time);
+ } while ( 0 );
+
+ if ( NULL == c )
+ set_status(ur, 0);
+
+ spin_unlock(&hd->vmport_lock);
+}
+
+static int hvm_set_guestinfo(struct vmport_state *vs,
+ struct xen_hvm_vmport_guest_info *a,
+ char *key, char *value)
+{
+ int idx;
+ int free_idx = -1;
+ int rc = 0;
+
+ for ( idx = 0; idx < vs->used_guestinfo; idx++ )
+ {
+ if ( !vs->guestinfo[idx] )
+ {
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING,
+ "idx=%d not allocated, but used_guestinfo=%d\n",
+ idx, vs->used_guestinfo);
+#endif
+ }
+ else if ( (vs->guestinfo[idx]->key_len == a->key_length) &&
+ (memcmp(key,
+ vs->guestinfo[idx]->key_data,
+ vs->guestinfo[idx]->key_len) == 0) )
+ {
+ vs->guestinfo[idx]->val_len = a->value_length;
+ memcpy(vs->guestinfo[idx]->val_data, value, a->value_length);
+ break;
+
+ }
+ else if ( (vs->guestinfo[idx]->key_len == 0) &&
+ (free_idx == -1) )
+ free_idx = idx;
+ }
+
+ if ( idx >= vs->used_guestinfo )
+ {
+ if ( free_idx == -1 )
+ rc = -EBUSY;
+ else
+ {
+ vs->guestinfo[free_idx]->key_len = a->key_length;
+ memcpy(vs->guestinfo[free_idx]->key_data, key, a->key_length);
+ vs->guestinfo[free_idx]->val_len = a->value_length;
+ memcpy(vs->guestinfo[free_idx]->val_data, value, a->value_length);
+ }
+ }
+
+ /* Delete any duplicate entry */
+ if ( rc == 0 )
+ hvm_del_guestinfo_jumbo(vs, key, a->key_length);
+
+ return rc;
+}
+
+static int hvm_set_guestinfo_jumbo(struct vmport_state *vs,
+ struct xen_hvm_vmport_guest_info *a,
+ char *key, char *value)
+{
+ int idx;
+ int free_idx = -1;
+ int rc = 0;
+
+ for ( idx = 0; idx < vs->used_guestinfo_jumbo; idx++ )
+ {
+
+ if ( !vs->guestinfo_jumbo[idx] )
+ {
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING,
+ "idx=%d not allocated, but used_guestinfo_jumbo=%d\n",
+ idx, vs->used_guestinfo_jumbo);
+#endif
+ }
+ else if ( (vs->guestinfo_jumbo[idx]->key_len == a->key_length) &&
+ (memcmp(key, vs->guestinfo_jumbo[idx]->key_data,
+ vs->guestinfo_jumbo[idx]->key_len) == 0) )
+ {
+ vs->guestinfo_jumbo[idx]->val_len = a->value_length;
+ memcpy(vs->guestinfo_jumbo[idx]->val_data, value, a->value_length);
+ break;
+
+ }
+ else if ( (vs->guestinfo_jumbo[idx]->key_len == 0) &&
+ (free_idx == -1) )
+ free_idx = idx;
+ }
+
+ if ( idx >= vs->used_guestinfo_jumbo )
+ {
+ if ( free_idx == -1 )
+ rc = -EBUSY;
+ else
+ {
+ vs->guestinfo_jumbo[free_idx]->key_len = a->key_length;
+ memcpy(vs->guestinfo_jumbo[free_idx]->key_data,
+ key, a->key_length);
+ vs->guestinfo_jumbo[free_idx]->val_len = a->value_length;
+ memcpy(vs->guestinfo_jumbo[free_idx]->val_data,
+ value, a->value_length);
+ }
+ }
+
+ /* Delete any duplicate entry */
+ if ( rc == 0 )
+ hvm_del_guestinfo(vs, key, a->key_length);
+
+ return rc;
+}
+
+static int hvm_get_guestinfo(struct vmport_state *vs,
+ struct xen_hvm_vmport_guest_info *a,
+ char *key, char *value)
+{
+ int idx;
+ int rc = 0;
+
+ if ( a->key_length == 0 )
+ {
+ /*
+ * Here we are iterating on getting all guestinfo entries
+ * using index
+ */
+ idx = a->value_length;
+ if ( idx >= vs->used_guestinfo ||
+ !vs->guestinfo[idx] )
+ rc = -ENOENT;
+ else
+ {
+ a->key_length = vs->guestinfo[idx]->key_len;
+ memcpy(a->data, vs->guestinfo[idx]->key_data, a->key_length);
+ a->value_length = vs->guestinfo[idx]->val_len;
+ memcpy(&a->data[a->key_length], vs->guestinfo[idx]->val_data,
+ a->value_length);
+ rc = 0;
+ }
+ }
+ else
+ {
+ for ( idx = 0; idx < vs->used_guestinfo; idx++ )
+ {
+ if ( vs->guestinfo[idx] &&
+ (vs->guestinfo[idx]->key_len == a->key_length) &&
+ (memcmp(key, vs->guestinfo[idx]->key_data,
+ vs->guestinfo[idx]->key_len) == 0) )
+ {
+ a->value_length = vs->guestinfo[idx]->val_len;
+ memcpy(value, vs->guestinfo[idx]->val_data,
+ a->value_length);
+ rc = 0;
+ break;
+ }
+ }
+ if ( idx >= vs->used_guestinfo )
+ rc = -ENOENT;
+ }
+ return rc;
+}
+
+static int hvm_get_guestinfo_jumbo(struct vmport_state *vs,
+ struct xen_hvm_vmport_guest_info *a,
+ char *key, char *value)
+{
+ int idx, total_entries;
+ int rc = 0;
+
+ if ( a->key_length == 0 )
+ {
+ /*
+ * Here we are iterating on getting all guestinfo entries
+ * using index
+ */
+ total_entries = vs->used_guestinfo + vs->used_guestinfo_jumbo;
+
+ /* Input index is in a->value_length */
+ if ( a->value_length >= total_entries )
+ {
+ rc = -ENOENT;
+ return rc;
+ }
+ idx = a->value_length - vs->used_guestinfo;
+ if ( idx >= vs->used_guestinfo_jumbo ||
+ !vs->guestinfo_jumbo[idx] )
+ rc = -ENOENT;
+ else
+ {
+ a->key_length = vs->guestinfo_jumbo[idx]->key_len;
+ memcpy(a->data, vs->guestinfo_jumbo[idx]->key_data, a->key_length);
+ a->value_length = vs->guestinfo_jumbo[idx]->val_len;
+ memcpy(&a->data[a->key_length],
+ vs->guestinfo_jumbo[idx]->val_data, a->value_length);
+ rc = 0;
+ }
+ }
+ else
+ {
+ for ( idx = 0; idx < vs->used_guestinfo_jumbo; idx++ )
+ {
+ if ( vs->guestinfo_jumbo[idx] &&
+ (vs->guestinfo_jumbo[idx]->key_len == a->key_length) &&
+ (memcmp(key, vs->guestinfo_jumbo[idx]->key_data,
+ vs->guestinfo_jumbo[idx]->key_len) == 0) )
+ {
+ a->value_length = vs->guestinfo_jumbo[idx]->val_len;
+ memcpy(value, vs->guestinfo_jumbo[idx]->val_data,
+ a->value_length);
+ rc = 0;
+ break;
+ }
+ }
+ if ( idx >= vs->used_guestinfo_jumbo )
+ rc = -ENOENT;
+ }
+ return rc;
+}
+
+int vmport_rpc_hvmop_precheck(unsigned long op,
+ struct xen_hvm_vmport_guest_info *a)
+{
+ int new_key_length = a->key_length;
+
+ if ( new_key_length > strlen("guestinfo.") )
+ {
+ if ( (size_t)new_key_length + (size_t)a->value_length >
+ sizeof(a->data) )
+ return -EINVAL;
+ if ( memcmp(a->data, "guestinfo.", strlen("guestinfo.")) == 0 )
+ new_key_length -= strlen("guestinfo.");
+ if ( new_key_length > VMPORT_MAX_KEY_LEN )
+ {
+ gdprintk(XENLOG_ERR, "bad key len %d\n", new_key_length);
+ return -EINVAL;
+ }
+ if ( a->value_length > VMPORT_MAX_VAL_JUMBO_LEN )
+ {
+ gdprintk(XENLOG_ERR, "bad val len %d\n", a->value_length);
+ return -EINVAL;
+ }
+ }
+ else if ( new_key_length > 0 )
+ {
+ if ( (size_t)new_key_length + (size_t)a->value_length >
+ sizeof(a->data) )
+ return -EINVAL;
+ if ( new_key_length > VMPORT_MAX_KEY_LEN )
+ {
+ gdprintk(XENLOG_ERR, "bad key len %d", new_key_length);
+ return -EINVAL;
+ }
+ if ( a->value_length > VMPORT_MAX_VAL_JUMBO_LEN )
+ {
+ gdprintk(XENLOG_ERR, "bad val len %d\n", a->value_length);
+ return -EINVAL;
+ }
+ }
+ else if ( (new_key_length == 0) && (op == HVMOP_set_vmport_guest_info) )
+ return -EINVAL;
+
+ return 0;
+}
+
+int vmport_rpc_hvmop_do(struct domain *d, unsigned long op,
+ struct xen_hvm_vmport_guest_info *a)
+{
+ char *key = NULL;
+ char *value = NULL;
+ struct vmport_state *vs = d->arch.hvm_domain.vmport_data;
+ int i, total_entries;
+ vmport_guestinfo_t *add_slots[5];
+ vmport_guestinfo_jumbo_t *add_slots_jumbo[2];
+ int num_slots = 0, num_slots_jumbo = 0, num_free_slots = 0;
+ int rc = 0;
+
+ if ( a->key_length > strlen("guestinfo.") )
+ {
+ if ( memcmp(a->data, "guestinfo.", strlen("guestinfo.")) == 0 )
+ {
+ key = &a->data[strlen("guestinfo.")];
+ a->key_length -= strlen("guestinfo.");
+ }
+ else
+ key = &a->data[0];
+ value = key + a->key_length;
+ }
+ else if ( a->key_length > 0 )
+ {
+ key = &a->data[0];
+ value = key + a->key_length;
+ }
+
+ total_entries = vs->used_guestinfo + vs->used_guestinfo_jumbo;
+
+ if ( (a->key_length == 0) && (a->value_length >= total_entries) )
+ {
+ /*
+ * When key length is zero, we are interating on
+ * get-guest-info hypercalls to retrieve all guestinfo
+ * entries using index passed in a->value_length
+ */
+ return -E2BIG;
+ }
+
+ num_free_slots = 0;
+ for ( i = 0; i < vs->used_guestinfo; i++ )
+ {
+ if ( vs->guestinfo[i] &&
+ (vs->guestinfo[i]->key_len == 0) )
+ num_free_slots++;
+ }
+ if ( num_free_slots < 5 )
+ {
+ num_slots = 5 - num_free_slots;
+ if ( vs->used_guestinfo + num_slots > VMPORT_MAX_NUM_KEY )
+ num_slots = VMPORT_MAX_NUM_KEY - vs->used_guestinfo;
+ for ( i = 0; i < num_slots; i++ )
+ add_slots[i] = xzalloc(vmport_guestinfo_t);
+ }
+
+ num_free_slots = 0;
+ for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
+ {
+ if ( vs->guestinfo_jumbo[i] &&
+ (vs->guestinfo_jumbo[i]->key_len == 0) )
+ num_free_slots++;
+ }
+ if ( num_free_slots < 1 )
+ {
+ num_slots_jumbo = 1 - num_free_slots;
+ if ( vs->used_guestinfo_jumbo + num_slots_jumbo >
+ VMPORT_MAX_NUM_JUMBO_KEY )
+ num_slots_jumbo = VMPORT_MAX_NUM_JUMBO_KEY -
+ vs->used_guestinfo_jumbo;
+ for ( i = 0; i < num_slots_jumbo; i++ )
+ add_slots_jumbo[i] = xzalloc(vmport_guestinfo_jumbo_t);
+ }
+
+ spin_lock(&d->arch.hvm_domain.vmport_lock);
+
+ for ( i = 0; i < num_slots; i++ )
+ vs->guestinfo[vs->used_guestinfo + i] = add_slots[i];
+ vs->used_guestinfo += num_slots;
+
+ for ( i = 0; i < num_slots_jumbo; i++ )
+ vs->guestinfo_jumbo[vs->used_guestinfo_jumbo + i] =
+ add_slots_jumbo[i];
+ vs->used_guestinfo_jumbo += num_slots_jumbo;
+
+ if ( op == HVMOP_set_vmport_guest_info )
+ {
+ if ( a->value_length > VMPORT_MAX_VAL_LEN )
+ rc = hvm_set_guestinfo_jumbo(vs, a, key, value);
+ else
+ rc = hvm_set_guestinfo(vs, a, key, value);
+ }
+ else
+ {
+ /* Get Guest Info */
+ rc = hvm_get_guestinfo(vs, a, key, value);
+ if ( rc != 0 )
+ rc = hvm_get_guestinfo_jumbo(vs, a, key, value);
+ }
+ spin_unlock(&d->arch.hvm_domain.vmport_lock);
+
+ return rc;
+}
+
+int vmport_rpc_init(struct domain *d)
+{
+ struct vmport_state *vs = xzalloc(struct vmport_state);
+ int i;
+
+ spin_lock_init(&d->arch.hvm_domain.vmport_lock);
+ d->arch.hvm_domain.vmport_data = vs;
+
+ if ( !vs )
+ return -ENOMEM;
+
+ /*
+ * Any value is fine here. In fact a random number may better.
+ * It is used to help validate that a both sides are talking
+ * about the same channel.
+ */
+ vs->open_cookie = 435;
+
+ vs->used_guestinfo = 10;
+ for ( i = 0; i < vs->used_guestinfo; i++ )
+ vs->guestinfo[i] = xzalloc(vmport_guestinfo_t);
+
+ vs->used_guestinfo_jumbo = 2;
+ for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
+ vs->guestinfo_jumbo[i] = xzalloc(vmport_guestinfo_jumbo_t);
+
+ vmport_flush(&d->arch.hvm_domain);
+
+ d->arch.hvm_domain.params[HVM_PARAM_VMPORT_RESET_TIME] = 14;
+
+ return 0;
+}
+
+void vmport_rpc_deinit(struct domain *d)
+{
+ struct vmport_state *vs = d->arch.hvm_domain.vmport_data;
+ int i;
+
+ if ( !vs )
+ return;
+
+ for ( i = 0; i < vs->used_guestinfo; i++ )
+ xfree(vs->guestinfo[i]);
+ for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
+ xfree(vs->guestinfo_jumbo[i]);
+ xfree(vs);
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-set-style: "BSD"
+ * c-basic-offset: 4
+ * tab-width: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/asm-x86/hvm/domain.h b/xen/include/asm-x86/hvm/domain.h
index 93b081b..8fa9090 100644
--- a/xen/include/asm-x86/hvm/domain.h
+++ b/xen/include/asm-x86/hvm/domain.h
@@ -107,6 +107,10 @@ struct hvm_domain {
/* emulated irq to pirq */
struct radix_tree_root emuirq_pirq;
+ /* VMware special port RPC */
+ spinlock_t vmport_lock;
+ struct vmport_state *vmport_data;
+
uint64_t *params;
/* Memory ranges with pinned cache attributes. */
diff --git a/xen/include/asm-x86/hvm/vmport.h b/xen/include/asm-x86/hvm/vmport.h
index 74a3b66..a5c70bb 100644
--- a/xen/include/asm-x86/hvm/vmport.h
+++ b/xen/include/asm-x86/hvm/vmport.h
@@ -25,6 +25,15 @@
#define VMPORT_LOG_VGP_UNKNOWN (1 << 3)
#define VMPORT_LOG_REALMODE_GP (1 << 4)
+#define VMPORT_LOG_RECV (1 << 8)
+#define VMPORT_LOG_SEND (1 << 9)
+#define VMPORT_LOG_SKIP_SEND (1 << 10)
+#define VMPORT_LOG_ERROR (1 << 11)
+
+#define VMPORT_LOG_SWEEP (1 << 12)
+#define VMPORT_LOG_PING (1 << 13)
+
+
extern unsigned int opt_vmport_debug;
#define VMPORT_DBG_LOG(level, _f, _a...) \
do { \
@@ -42,6 +51,13 @@ int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val);
int vmport_gp_check(struct cpu_user_regs *regs, struct vcpu *v,
unsigned long inst_len, unsigned long inst_addr,
unsigned long ei1, unsigned long ei2);
+int vmport_rpc_init(struct domain *d);
+void vmport_rpc_deinit(struct domain *d);
+void vmport_rpc(struct hvm_domain *hd, struct cpu_user_regs *ur);
+int vmport_rpc_hvmop_precheck(unsigned long op,
+ struct xen_hvm_vmport_guest_info *a);
+int vmport_rpc_hvmop_do(struct domain *d, unsigned long op,
+ struct xen_hvm_vmport_guest_info *a);
#endif /* ASM_X86_HVM_VMPORT_H__ */
diff --git a/xen/include/public/arch-x86/hvm/vmporttype.h b/xen/include/public/arch-x86/hvm/vmporttype.h
new file mode 100644
index 0000000..98875d2
--- /dev/null
+++ b/xen/include/public/arch-x86/hvm/vmporttype.h
@@ -0,0 +1,118 @@
+/*
+ * vmporttype.h: HVM VMPORT structure definitions
+ *
+ *
+ * Copyright (C) 2012 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef ASM_X86_HVM_VMPORTTYPE_H__
+#define ASM_X86_HVM_VMPORTTYPE_H__
+
+#define VMPORT_MAX_KEY_LEN 30
+#define VMPORT_MAX_VAL_LEN 128
+#define VMPORT_MAX_NUM_KEY 64
+#define VMPORT_MAX_NUM_JUMBO_KEY 4
+#define VMPORT_MAX_VAL_JUMBO_LEN 4096
+
+#define VMPORT_MAX_SEND_BUF ((22 + VMPORT_MAX_KEY_LEN + \
+ VMPORT_MAX_VAL_JUMBO_LEN + 3)/4)
+#define VMPORT_MAX_RECV_BUF ((2 + VMPORT_MAX_VAL_LEN + 3)/4)
+#define VMPORT_MAX_RECV_JUMBO_BUF ((2 + VMPORT_MAX_VAL_JUMBO_LEN + 3)/4)
+#define VMPORT_MAX_CHANS 6
+#define VMPORT_MAX_BKTS 8
+
+#define VMPORT_SAVE_VERSION 0xabcd0001
+
+typedef struct
+{
+ uint8_t key_len;
+ uint8_t val_len;
+ char key_data[VMPORT_MAX_KEY_LEN];
+ char val_data[VMPORT_MAX_VAL_LEN];
+} vmport_guestinfo_t;
+
+typedef struct
+{
+ uint16_t val_len;
+ uint8_t key_len;
+ char key_data[VMPORT_MAX_KEY_LEN];
+ char val_data[VMPORT_MAX_VAL_JUMBO_LEN];
+} vmport_guestinfo_jumbo_t;
+
+typedef struct __attribute__((packed))
+{
+ uint16_t recv_len;
+ uint16_t recv_idx;
+}
+vmport_bucket_control_t;
+
+typedef struct __attribute__((packed))
+{
+ vmport_bucket_control_t ctl;
+ uint32_t recv_buf[VMPORT_MAX_RECV_BUF + 1];
+}
+vmport_bucket_t;
+
+typedef struct __attribute__((packed))
+{
+ vmport_bucket_control_t ctl;
+ uint32_t recv_buf[VMPORT_MAX_RECV_JUMBO_BUF + 1];
+}
+vmport_jumbo_bucket_t;
+
+typedef struct __attribute__((packed))
+{
+ uint64_t active_time;
+ uint32_t chan_id;
+ uint32_t cookie;
+ uint32_t proto_num;
+ uint16_t send_len;
+ uint16_t send_idx;
+ uint8_t jumbo;
+ uint8_t recv_read;
+ uint8_t recv_write;
+ uint8_t recv_chan_pad[1];
+}
+vmport_channel_control_t;
+
+typedef struct __attribute__((packed))
+{
+ vmport_channel_control_t ctl;
+ vmport_bucket_t recv_bkt[VMPORT_MAX_BKTS];
+ vmport_jumbo_bucket_t jumbo_recv_bkt;
+ uint32_t send_buf[VMPORT_MAX_SEND_BUF + 1];
+}
+vmport_channel_t;
+
+struct vmport_state
+{
+ uint64_t ping_time;
+ uint32_t open_cookie;
+ uint32_t used_guestinfo;
+ uint32_t used_guestinfo_jumbo;
+ uint8_t max_chans;
+ uint8_t state_pad[3];
+ vmport_channel_t chans[VMPORT_MAX_CHANS];
+ vmport_guestinfo_t *guestinfo[VMPORT_MAX_NUM_KEY];
+ vmport_guestinfo_jumbo_t *guestinfo_jumbo[VMPORT_MAX_NUM_JUMBO_KEY];
+};
+
+#endif /* ASM_X86_HVM_VMPORTTYPE_H__ */
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
diff --git a/xen/include/public/hvm/hvm_op.h b/xen/include/public/hvm/hvm_op.h
index eeb0a60..8e1e072 100644
--- a/xen/include/public/hvm/hvm_op.h
+++ b/xen/include/public/hvm/hvm_op.h
@@ -369,6 +369,24 @@ DEFINE_XEN_GUEST_HANDLE(xen_hvm_set_ioreq_server_state_t);
#endif /* defined(__XEN__) || defined(__XEN_TOOLS__) */
+/* Get/set vmport subcommands */
+#define HVMOP_get_vmport_guest_info 23
+#define HVMOP_set_vmport_guest_info 24
+#define VMPORT_GUEST_INFO_KEY_MAX 40
+#define VMPORT_GUEST_INFO_VAL_MAX 4096
+struct xen_hvm_vmport_guest_info {
+ /* Domain to be accessed */
+ domid_t domid;
+ /* key length */
+ uint16_t key_length;
+ /* value length */
+ uint16_t value_length;
+ /* key and value data */
+ char data[VMPORT_GUEST_INFO_KEY_MAX + VMPORT_GUEST_INFO_VAL_MAX];
+};
+typedef struct xen_hvm_vmport_guest_info xen_hvm_vmport_guest_info_t;
+DEFINE_XEN_GUEST_HANDLE(xen_hvm_vmport_guest_info_t);
+
#endif /* __XEN_PUBLIC_HVM_HVM_OP_H__ */
/*
diff --git a/xen/include/public/hvm/params.h b/xen/include/public/hvm/params.h
index dee6d68..722e30b 100644
--- a/xen/include/public/hvm/params.h
+++ b/xen/include/public/hvm/params.h
@@ -153,7 +153,10 @@
/* Params for VMware */
#define HVM_PARAM_VMWARE_HW 35
+#define HVM_PARAM_VMPORT_BUILD_NUMBER_TIME 36
+#define HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE 37
+#define HVM_PARAM_VMPORT_RESET_TIME 38
-#define HVM_NR_PARAMS 36
+#define HVM_NR_PARAMS 39
#endif /* __XEN_PUBLIC_HVM_PARAMS_H__ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 09/16] tools: Add limited support of VMware's hyper-call rpc
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (7 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 10/16] Add VMware tool's triggers Don Slutz
` (6 subsequent siblings)
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
This guestinfo support is provided via libxc. libxl support has
not be written.
Note: VMware RPC support is only available on HVM domU.
This interface is an extension of __HYPERVISOR_HVM_op. It was
picked because xc_get_hvm_param() also uses it and VMware guest
info is a lot like a hvm param.
The HVMOP_get_vmport_guest_info is used by two libxc functions,
xc_get_vmport_guest_info and xc_fetch_all_vmport_guest_info.
xc_fetch_all_vmport_guest_info is designed to be used to fetch all
currently set guestinfo values.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
tools/libxc/xc_domain.c | 115 ++++++++++++++++++++++++++++++++++++++++++++++++
tools/libxc/xenctrl.h | 24 ++++++++++
2 files changed, 139 insertions(+)
diff --git a/tools/libxc/xc_domain.c b/tools/libxc/xc_domain.c
index 1a6f90a..ce24dad 100644
--- a/tools/libxc/xc_domain.c
+++ b/tools/libxc/xc_domain.c
@@ -1577,6 +1577,121 @@ int xc_hvm_set_ioreq_server_state(xc_interface *xch,
return rc;
}
+int xc_set_vmport_guest_info(xc_interface *handle,
+ domid_t dom,
+ unsigned int key_len,
+ char *key,
+ unsigned int val_len,
+ char *val)
+{
+ DECLARE_HYPERCALL;
+ DECLARE_HYPERCALL_BUFFER(xen_hvm_vmport_guest_info_t, arg);
+ int rc;
+
+ if ( (key_len < 1) ||
+ (key_len > VMPORT_GUEST_INFO_KEY_MAX) ||
+ (val_len > VMPORT_GUEST_INFO_VAL_MAX) )
+ return -1;
+
+ arg = xc_hypercall_buffer_alloc(handle, arg, sizeof(*arg));
+ if ( arg == NULL )
+ return -1;
+
+ hypercall.op = __HYPERVISOR_hvm_op;
+ hypercall.arg[0] = HVMOP_set_vmport_guest_info;
+ hypercall.arg[1] = HYPERCALL_BUFFER_AS_ARG(arg);
+ arg->domid = dom;
+ arg->key_length = key_len;
+ arg->value_length = val_len;
+ memcpy(arg->data, key, key_len);
+ memcpy(&arg->data[key_len], val, val_len);
+ rc = do_xen_hypercall(handle, &hypercall);
+ xc_hypercall_buffer_free(handle, arg);
+ return rc;
+}
+
+int xc_get_vmport_guest_info(xc_interface *handle,
+ domid_t dom,
+ unsigned int key_len,
+ char *key,
+ unsigned int val_max,
+ unsigned int *val_len,
+ char *val)
+{
+ DECLARE_HYPERCALL;
+ DECLARE_HYPERCALL_BUFFER(xen_hvm_vmport_guest_info_t, arg);
+ int rc;
+
+ if ( (key_len < 1) ||
+ (key_len > VMPORT_GUEST_INFO_KEY_MAX) )
+ return -1;
+
+ arg = xc_hypercall_buffer_alloc(handle, arg, sizeof(*arg));
+
+ hypercall.op = __HYPERVISOR_hvm_op;
+ hypercall.arg[0] = HVMOP_get_vmport_guest_info;
+ hypercall.arg[1] = HYPERCALL_BUFFER_AS_ARG(arg);
+ arg->domid = dom;
+ arg->key_length = key_len;
+ arg->value_length = 0;
+ *val_len = 0;
+ memcpy(arg->data, key, key_len);
+ rc = do_xen_hypercall(handle, &hypercall);
+ if ( rc == 0 )
+ {
+ *val_len = arg->value_length;
+ if ( arg->value_length > val_max )
+ arg->value_length = val_max;
+ memcpy(val, &arg->data[key_len], arg->value_length);
+ }
+ xc_hypercall_buffer_free(handle, arg);
+ return rc;
+}
+
+int xc_fetch_all_vmport_guest_info(xc_interface *handle,
+ domid_t dom,
+ unsigned int idx,
+ unsigned int key_max,
+ unsigned int *key_len,
+ char *key,
+ unsigned int val_max,
+ unsigned int *val_len,
+ char *val)
+{
+ DECLARE_HYPERCALL;
+ DECLARE_HYPERCALL_BUFFER(xen_hvm_vmport_guest_info_t, arg);
+ int rc;
+
+ arg = xc_hypercall_buffer_alloc(handle, arg, sizeof(*arg));
+
+ hypercall.op = __HYPERVISOR_hvm_op;
+ hypercall.arg[0] = HVMOP_get_vmport_guest_info;
+ hypercall.arg[1] = HYPERCALL_BUFFER_AS_ARG(arg);
+ arg->domid = dom;
+ arg->key_length = 0;
+ arg->value_length = idx;
+ *key_len = 0;
+ *val_len = 0;
+ rc = do_xen_hypercall(handle, &hypercall);
+ if ( rc == 0 )
+ {
+ uint16_t val_off = arg->key_length;
+
+ *key_len = arg->key_length;
+ if ( arg->key_length > key_max )
+ arg->key_length = key_max;
+ memcpy(key, arg->data, arg->key_length);
+ *val_len = arg->value_length;
+ if ( arg->value_length > val_max )
+ arg->value_length = val_max;
+ memcpy(val,
+ &arg->data[val_off],
+ arg->value_length);
+ }
+ xc_hypercall_buffer_free(handle, arg);
+ return rc;
+}
+
int xc_domain_setdebugging(xc_interface *xch,
uint32_t domid,
unsigned int enable)
diff --git a/tools/libxc/xenctrl.h b/tools/libxc/xenctrl.h
index 1c8aa42..49d8d1e 100644
--- a/tools/libxc/xenctrl.h
+++ b/tools/libxc/xenctrl.h
@@ -2019,6 +2019,30 @@ int xc_hvm_destroy_ioreq_server(xc_interface *xch,
domid_t domid,
ioservid_t id);
+int xc_set_vmport_guest_info(xc_interface *handle,
+ domid_t dom,
+ unsigned int key_len,
+ char *key,
+ unsigned int val_len,
+ char *val);
+int xc_get_vmport_guest_info(xc_interface *handle,
+ domid_t dom,
+ unsigned int key_len,
+ char *key,
+ unsigned int val_max,
+ unsigned int *val_len,
+ char *val);
+int xc_fetch_all_vmport_guest_info(xc_interface *handle,
+ domid_t dom,
+ unsigned int idx,
+ unsigned int key_max,
+ unsigned int *key_len,
+ char *key,
+ unsigned int val_max,
+ unsigned int *val_len,
+ char *val);
+
+
/* HVM guest pass-through */
int xc_assign_device(xc_interface *xch,
uint32_t domid,
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 10/16] Add VMware tool's triggers
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (8 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 09/16] tools: " Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 11/16] Add live migration of VMware's hyper-call RPC Don Slutz
` (5 subsequent siblings)
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
These are not the same as the PV control interface; they
are more like the ACPI power event.
For anything to happen when they are used, the domU must
be running a VMware tools daemon that is polling for triggers.
If the domU is running VMware tools, then the "build version" of
the tools is also available via xc_get_HVM_param(). This also
enables the use of new triggers that will use the VMware hyper-call
to do some limited control of the domU. The most useful are
poweroff and reboot. Since a guest process needs to be running
for these to work, a tool stack should check that the build version
is non zero before assuming these will work.
The 2 hvm param's HVM_PARAM_VMPORT_BUILD_NUMBER_TIME and
HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE are how "build version" is
accessed. These 2 params are only allowed to be set to zero. The
HVM_PARAM_VMPORT_BUILD_NUMBER_TIME can be used to track the last
time the VMware tools in the guest responded. One such use would
be the health of the tools in the guest. The hvm param
HVM_PARAM_VMPORT_RESET_TIME controls how often to request them in
seconds minus 1. The minus 1 is to handle to 0 case. I.E. the
fastest that can be selected is every second. The default is 4
times a minute.
XEN_DOMCTL_SENDTRIGGER_VTPING is the same as what is done using
HVM_PARAM_VMPORT_RESET_TIME. This trigger allows the tool
stack to request a sooner update of HVM_PARAM_VMPORT_BUILD_NUMBER_TIME
and HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
xen/arch/x86/domctl.c | 34 ++++++++++++++++++++++++++++++++++
xen/include/asm-x86/hvm/vmport.h | 1 +
xen/include/public/domctl.h | 3 +++
3 files changed, 38 insertions(+)
diff --git a/xen/arch/x86/domctl.c b/xen/arch/x86/domctl.c
index 7a5de43..50596a6 100644
--- a/xen/arch/x86/domctl.c
+++ b/xen/arch/x86/domctl.c
@@ -23,6 +23,7 @@
#include <xen/paging.h>
#include <asm/irq.h>
#include <asm/hvm/hvm.h>
+#include <asm/hvm/vmport.h>
#include <asm/hvm/support.h>
#include <asm/hvm/cacheattr.h>
#include <asm/processor.h>
@@ -579,6 +580,39 @@ long arch_do_domctl(
}
break;
+ case XEN_DOMCTL_SENDTRIGGER_VTPOWER:
+ {
+ ret = -EINVAL;
+ if ( is_hvm_domain(d) )
+ {
+ ret = 0;
+ vmport_ctrl_send(&d->arch.hvm_domain, "OS_Halt");
+ }
+ }
+ break;
+
+ case XEN_DOMCTL_SENDTRIGGER_VTREBOOT:
+ {
+ ret = -EINVAL;
+ if ( is_hvm_domain(d) )
+ {
+ ret = 0;
+ vmport_ctrl_send(&d->arch.hvm_domain, "OS_Reboot");
+ }
+ }
+ break;
+
+ case XEN_DOMCTL_SENDTRIGGER_VTPING:
+ {
+ ret = -EINVAL;
+ if ( is_hvm_domain(d) )
+ {
+ ret = 0;
+ vmport_ctrl_send(&d->arch.hvm_domain, "ping");
+ }
+ }
+ break;
+
default:
ret = -ENOSYS;
}
diff --git a/xen/include/asm-x86/hvm/vmport.h b/xen/include/asm-x86/hvm/vmport.h
index a5c70bb..981e43a 100644
--- a/xen/include/asm-x86/hvm/vmport.h
+++ b/xen/include/asm-x86/hvm/vmport.h
@@ -58,6 +58,7 @@ int vmport_rpc_hvmop_precheck(unsigned long op,
struct xen_hvm_vmport_guest_info *a);
int vmport_rpc_hvmop_do(struct domain *d, unsigned long op,
struct xen_hvm_vmport_guest_info *a);
+void vmport_ctrl_send(struct hvm_domain *hd, char *msg);
#endif /* ASM_X86_HVM_VMPORT_H__ */
diff --git a/xen/include/public/domctl.h b/xen/include/public/domctl.h
index 7a0f691..2d9a534 100644
--- a/xen/include/public/domctl.h
+++ b/xen/include/public/domctl.h
@@ -462,6 +462,9 @@ DEFINE_XEN_GUEST_HANDLE(xen_domctl_address_size_t);
#define XEN_DOMCTL_SENDTRIGGER_INIT 2
#define XEN_DOMCTL_SENDTRIGGER_POWER 3
#define XEN_DOMCTL_SENDTRIGGER_SLEEP 4
+#define XEN_DOMCTL_SENDTRIGGER_VTPOWER 5
+#define XEN_DOMCTL_SENDTRIGGER_VTREBOOT 6
+#define XEN_DOMCTL_SENDTRIGGER_VTPING 7
struct xen_domctl_sendtrigger {
uint32_t trigger; /* IN */
uint32_t vcpu; /* IN */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 11/16] Add live migration of VMware's hyper-call RPC
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (9 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 10/16] Add VMware tool's triggers Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-12 13:54 ` Boris Ostrovsky
2014-09-11 18:36 ` [PATCH v4 12/16] Add dump of HVM_SAVE_CODE(VMPORT) to xen-hvmctx Don Slutz
` (4 subsequent siblings)
15 siblings, 1 reply; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
The VMware's hyper-call state is included in live migration and
save/restore. Because the max size of the VMware guestinfo is
large, then data is compressed and expanded in the
vmport_save_domain_ctxt and vmport_load_domain_ctxt.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
xen/arch/x86/hvm/vmware/vmport_rpc.c | 302 ++++++++++++++++++++++++++++++++-
xen/include/public/arch-x86/hvm/save.h | 39 ++++-
2 files changed, 339 insertions(+), 2 deletions(-)
diff --git a/xen/arch/x86/hvm/vmware/vmport_rpc.c b/xen/arch/x86/hvm/vmware/vmport_rpc.c
index 695e58f..e25ed09 100644
--- a/xen/arch/x86/hvm/vmware/vmport_rpc.c
+++ b/xen/arch/x86/hvm/vmware/vmport_rpc.c
@@ -39,7 +39,8 @@
#include <asm/hvm/vmport.h>
#include <asm/hvm/trace.h>
-#include "public/arch-x86/hvm/vmporttype.h"
+#include <public/hvm/save.h>
+#include <public/arch-x86/hvm/save.h>
#include "backdoor_def.h"
#include "guest_msg_def.h"
@@ -1270,6 +1271,305 @@ void vmport_rpc_deinit(struct domain *d)
xfree(vs);
}
+/* save and restore functions */
+
+static int vmport_save_domain_ctxt(struct domain *d, hvm_domain_context_t *h)
+{
+ struct hvm_vmport_context *ctxt;
+ struct hvm_save_descriptor *desc;
+ struct hvm_domain *hd = &d->arch.hvm_domain;
+ struct vmport_state *vs = hd->vmport_data;
+ char *p;
+ unsigned int guestinfo_size = 0;
+ unsigned int used_guestinfo = 0;
+ unsigned int used_guestinfo_jumbo = 0;
+ unsigned int chans_size;
+ unsigned int i;
+
+ /* Customized handling for entry since our entry is of variable length */
+ desc = (struct hvm_save_descriptor *)&h->data[h->cur];
+ if ( _hvm_init_entry(h, HVM_SAVE_CODE(VMPORT), 0,
+ HVM_SAVE_LENGTH(VMPORT)) )
+ return 1;
+ ctxt = (struct hvm_vmport_context *)&h->data[h->cur];
+
+ spin_lock(&hd->vmport_lock);
+
+ ctxt->version = VMPORT_SAVE_VERSION;
+ ctxt->ping_time = vs->ping_time;
+ ctxt->open_cookie = vs->open_cookie;
+ ctxt->used_guestinfo = vs->used_guestinfo;
+ ctxt->used_guestinfo_jumbo = vs->used_guestinfo_jumbo;
+
+ p = ctxt->u.packed.packed_data;
+
+ for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
+ {
+ unsigned int j;
+ unsigned int buf_max;
+
+ ctxt->u.packed.chan_ctl[i].chan = vs->chans[i].ctl;
+ buf_max = vs->chans[i].ctl.send_len;
+ if ( buf_max > VMPORT_MAX_SEND_BUF * 4 )
+ buf_max = VMPORT_MAX_SEND_BUF * 4;
+ memcpy(p, vs->chans[i].send_buf, buf_max);
+ p += buf_max;
+ for ( j = 0; j < VMPORT_MAX_BKTS; j++ )
+ {
+ ctxt->u.packed.chan_ctl[i].recv[j] = vs->chans[i].recv_bkt[j].ctl;
+ buf_max = vs->chans[i].recv_bkt[j].ctl.recv_len;
+ if ( buf_max > VMPORT_MAX_RECV_BUF * 4 )
+ buf_max = VMPORT_MAX_RECV_BUF * 4;
+ memcpy(p, vs->chans[i].recv_bkt[j].recv_buf, buf_max);
+ p += buf_max;
+ }
+ ctxt->u.packed.chan_ctl[i].jumbo = vs->chans[i].jumbo_recv_bkt.ctl;
+ buf_max = vs->chans[i].jumbo_recv_bkt.ctl.recv_len;
+ if ( buf_max > VMPORT_MAX_RECV_JUMBO_BUF * 4 )
+ buf_max = VMPORT_MAX_RECV_JUMBO_BUF * 4;
+ memcpy(p, vs->chans[i].jumbo_recv_bkt.recv_buf, buf_max);
+ p += buf_max;
+ }
+
+ chans_size = p - ctxt->u.packed.packed_data;
+
+ for ( i = 0; i < ctxt->used_guestinfo; i++ )
+ {
+ vmport_guestinfo_t *vg = vs->guestinfo[i];
+
+ if ( vg && vg->key_len )
+ {
+ guestinfo_size += sizeof(vg->key_len) + sizeof(vg->val_len) +
+ vg->key_len + vg->val_len;
+ used_guestinfo++;
+ ASSERT(sizeof(vg->key_len) == 1);
+ *p++ = (char) vg->key_len;
+ ASSERT(sizeof(vg->val_len) == 1);
+ *p++ = (char) vg->val_len;
+ if ( vg->key_len )
+ {
+ memcpy(p, vg->key_data, vg->key_len);
+ p += vg->key_len;
+ if ( vg->val_len )
+ {
+ memcpy(p, vg->val_data, vg->val_len);
+ p += vg->val_len;
+ }
+ }
+ }
+ }
+ ctxt->used_guestinfo = used_guestinfo;
+
+ for ( i = 0; i < ctxt->used_guestinfo_jumbo; i++ )
+ {
+ vmport_guestinfo_jumbo_t *vgj =
+ vs->guestinfo_jumbo[i];
+ if ( vgj && vgj->key_len )
+ {
+ guestinfo_size += sizeof(vgj->key_len) + sizeof(vgj->val_len) +
+ vgj->key_len + vgj->val_len;
+ used_guestinfo_jumbo++;
+ ASSERT(sizeof(vgj->key_len) == 1);
+ *p++ = (char) vgj->key_len;
+ memcpy(p, &vgj->val_len, sizeof(vgj->val_len));
+ p += sizeof(vgj->val_len);
+ if ( vgj->key_len )
+ {
+ memcpy(p, vgj->key_data, vgj->key_len);
+ p += vgj->key_len;
+ if ( vgj->val_len )
+ {
+ memcpy(p, vgj->val_data, vgj->val_len);
+ p += vgj->val_len;
+ }
+ }
+ }
+ }
+ ctxt->used_guestinfo_jumbo = used_guestinfo_jumbo;
+
+ ctxt->used_guestsize = guestinfo_size;
+
+ spin_unlock(&hd->vmport_lock);
+
+#ifndef NDEBUG
+ gdprintk(XENLOG_WARNING, "chans_size=%d guestinfo_size=%d, used=%ld\n",
+ chans_size, guestinfo_size,
+ p - ctxt->u.packed.packed_data);
+#endif
+ ASSERT(p - ctxt->u.packed.packed_data == chans_size + guestinfo_size);
+ ASSERT(desc->length >= p - (char *)ctxt);
+ desc->length = p - (char *)ctxt; /* Fixup length to be right */
+ h->cur += desc->length; /* Do _hvm_write_entry */
+ ASSERT(guestinfo_size < desc->length);
+
+ return 0;
+}
+
+static int vmport_load_domain_ctxt(struct domain *d, hvm_domain_context_t *h)
+{
+ struct hvm_vmport_context *ctxt;
+ struct hvm_save_descriptor *desc;
+ struct hvm_domain *hd = &d->arch.hvm_domain;
+ struct vmport_state *vs = hd->vmport_data;
+ unsigned int i;
+ uint8_t key_len;
+ uint16_t val_len;
+ char *p;
+ vmport_guestinfo_t *vg;
+ vmport_guestinfo_jumbo_t *vgj;
+ unsigned int loop_cnt;
+ unsigned int guestinfo_size;
+ unsigned int used_guestinfo;
+ unsigned int used_guestinfo_jumbo;
+
+ if ( !vs )
+ return -ENOMEM;
+
+ /* Customized checking for entry since our entry is of variable length */
+ desc = (struct hvm_save_descriptor *)&h->data[h->cur];
+ if ( sizeof(*desc) > h->size - h->cur )
+ {
+ printk(XENLOG_G_WARNING
+ "HVM%d restore: not enough data left to read descriptor"
+ "for type %lu\n", d->domain_id,
+ HVM_SAVE_CODE(VMPORT));
+ return -1;
+ }
+ if ( desc->length + sizeof(*desc) > h->size - h->cur )
+ {
+ printk(XENLOG_G_WARNING
+ "HVM%d restore: not enough data left to read %u bytes "
+ "for type %lu\n", d->domain_id, desc->length,
+ HVM_SAVE_CODE(VMPORT));
+ return -1;
+ }
+ if ( HVM_SAVE_CODE(VMPORT) != desc->typecode ||
+ (desc->length > HVM_SAVE_LENGTH(VMPORT)) )
+ {
+ printk(XENLOG_G_WARNING
+ "HVM%d restore mismatch: expected type %lu with max length %lu, "
+ "saw type %u length %u\n", d->domain_id, HVM_SAVE_CODE(VMPORT),
+ HVM_SAVE_LENGTH(VMPORT),
+ desc->typecode, desc->length);
+ return -1;
+ }
+ h->cur += sizeof(*desc);
+ /* Checking finished */
+
+ ctxt = (struct hvm_vmport_context *)&h->data[h->cur];
+ h->cur += desc->length;
+
+ if ( ctxt->version != VMPORT_SAVE_VERSION )
+ return -EINVAL;
+
+ spin_lock(&hd->vmport_lock);
+
+ vs->ping_time = ctxt->ping_time;
+ vs->open_cookie = ctxt->open_cookie;
+ vs->used_guestinfo = ctxt->used_guestinfo;
+ vs->used_guestinfo_jumbo = ctxt->used_guestinfo_jumbo;
+ guestinfo_size = ctxt->used_guestsize;
+ used_guestinfo = ctxt->used_guestinfo;
+ used_guestinfo_jumbo = ctxt->used_guestinfo_jumbo;
+
+ p = ctxt->u.packed.packed_data;
+
+ for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
+ {
+ unsigned int j;
+
+ vs->chans[i].ctl = ctxt->u.packed.chan_ctl[i].chan;
+ memcpy(vs->chans[i].send_buf, p, vs->chans[i].ctl.send_len);
+ p += vs->chans[i].ctl.send_len;
+ for ( j = 0; j < VMPORT_MAX_BKTS; j++ )
+ {
+ vs->chans[i].recv_bkt[j].ctl = ctxt->u.packed.chan_ctl[i].recv[j];
+ memcpy(vs->chans[i].recv_bkt[j].recv_buf, p,
+ vs->chans[i].recv_bkt[j].ctl.recv_len);
+ p += vs->chans[i].recv_bkt[j].ctl.recv_len;
+ }
+ vs->chans[i].jumbo_recv_bkt.ctl = ctxt->u.packed.chan_ctl[i].jumbo;
+ memcpy(vs->chans[i].jumbo_recv_bkt.recv_buf, p,
+ vs->chans[i].jumbo_recv_bkt.ctl.recv_len);
+ p += vs->chans[i].jumbo_recv_bkt.ctl.recv_len;
+ }
+
+
+ /* keep at least 10 total and 5 empty entries */
+ loop_cnt = (vs->used_guestinfo + 5) > 10 ?
+ (vs->used_guestinfo + 5) : 10;
+ for ( i = 0; i < loop_cnt; i++ )
+ {
+ if ( !vs->guestinfo[i] )
+ {
+ vs->guestinfo[i] = xzalloc(vmport_guestinfo_t);
+ }
+ if ( i < vs->used_guestinfo
+ && guestinfo_size > 0 )
+ {
+ key_len = (uint8_t)*p++;
+ val_len = (uint8_t)*p++;
+ guestinfo_size -= 2;
+ if ( guestinfo_size >= key_len + val_len )
+ {
+ vg = vs->guestinfo[i];
+ if ( key_len )
+ {
+ vg->key_len = key_len;
+ vg->val_len = val_len;
+ memcpy(vg->key_data, p, key_len);
+ p += key_len;
+ memcpy(vg->val_data, p, val_len);
+ p += val_len;
+ guestinfo_size -= key_len + val_len;
+ }
+ }
+ }
+ }
+ vs->used_guestinfo = loop_cnt;
+
+ /* keep at least 2 total and 1 empty entries */
+ loop_cnt = (vs->used_guestinfo_jumbo + 1) > 2 ?
+ (vs->used_guestinfo_jumbo + 1) : 2;
+ for ( i = 0; i < loop_cnt; i++ )
+ {
+ if ( !vs->guestinfo_jumbo[i] )
+ {
+ vs->guestinfo_jumbo[i] = xzalloc(vmport_guestinfo_jumbo_t);
+ }
+ if ( i < vs->used_guestinfo_jumbo
+ && guestinfo_size > 0 )
+ {
+ key_len = (uint8_t)*p++;
+ memcpy(&val_len, p, 2);
+ p += 2;
+ guestinfo_size -= 3;
+ if ( guestinfo_size >= key_len + val_len )
+ {
+ vgj = vs->guestinfo_jumbo[i];
+ if ( key_len )
+ {
+ vgj->key_len = key_len;
+ vgj->val_len = val_len;
+ memcpy(vgj->key_data, p, key_len);
+ p += key_len;
+ memcpy(vgj->val_data, p, val_len);
+ p += val_len;
+ guestinfo_size -= key_len + val_len;
+ }
+ }
+ }
+ }
+ vs->used_guestinfo_jumbo = loop_cnt;
+
+ spin_unlock(&hd->vmport_lock);
+
+ return 0;
+}
+
+HVM_REGISTER_SAVE_RESTORE(VMPORT, vmport_save_domain_ctxt,
+ vmport_load_domain_ctxt, 1, HVMSR_PER_DOM);
+
/*
* Local variables:
* mode: C
diff --git a/xen/include/public/arch-x86/hvm/save.h b/xen/include/public/arch-x86/hvm/save.h
index 16d85a3..7cdcb8f 100644
--- a/xen/include/public/arch-x86/hvm/save.h
+++ b/xen/include/public/arch-x86/hvm/save.h
@@ -26,6 +26,8 @@
#ifndef __XEN_PUBLIC_HVM_SAVE_X86_H__
#define __XEN_PUBLIC_HVM_SAVE_X86_H__
+#include "vmporttype.h"
+
/*
* Save/restore header: general info about the save file.
*/
@@ -610,9 +612,44 @@ struct hvm_msr {
#define CPU_MSR_CODE 20
+/*
+ * VMware context.
+ */
+struct hvm_vmport_context {
+ uint32_t version;
+ uint32_t used_guestsize;
+ uint64_t ping_time;
+ uint32_t open_cookie;
+ uint32_t used_guestinfo;
+ uint32_t used_guestinfo_jumbo;
+ union {
+ struct {
+ vmport_channel_t chans[VMPORT_MAX_CHANS];
+ vmport_guestinfo_t guestinfo[VMPORT_MAX_NUM_KEY];
+ vmport_guestinfo_jumbo_t guestinfo_jumbo[VMPORT_MAX_NUM_JUMBO_KEY];
+ } full;
+ struct {
+ struct {
+ vmport_channel_control_t chan;
+ vmport_bucket_control_t recv[VMPORT_MAX_BKTS];
+ vmport_bucket_control_t jumbo;
+ } chan_ctl[VMPORT_MAX_CHANS];
+#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
+ char packed_data[];
+#elif defined(__GNUC__)
+ char packed_data[0];
+#else
+ char packed_data[1 /* Variable length */];
+#endif
+ } packed;
+ } u;
+};
+
+DECLARE_HVM_SAVE_TYPE(VMPORT, 21, struct hvm_vmport_context);
+
/*
* Largest type-code in use
*/
-#define HVM_SAVE_CODE_MAX 20
+#define HVM_SAVE_CODE_MAX 21
#endif /* __XEN_PUBLIC_HVM_SAVE_X86_H__ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 12/16] Add dump of HVM_SAVE_CODE(VMPORT) to xen-hvmctx.
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (10 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 11/16] Add live migration of VMware's hyper-call RPC Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 13/16] Add xen-hvm-param Don Slutz
` (3 subsequent siblings)
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
It also does some validation of the compressed data. Currently expects
that all guest info are printable strings.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
tools/misc/xen-hvmctx.c | 229 ++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 229 insertions(+)
diff --git a/tools/misc/xen-hvmctx.c b/tools/misc/xen-hvmctx.c
index 5a69245..d076091 100644
--- a/tools/misc/xen-hvmctx.c
+++ b/tools/misc/xen-hvmctx.c
@@ -399,6 +399,225 @@ static void dump_tsc_adjust(void)
printf(" TSC_ADJUST: tsc_adjust %" PRIx64 "\n", p.tsc_adjust);
}
+static void dump_vmport(int vmport_size)
+{
+ int i;
+ HVM_SAVE_TYPE(VMPORT) *vp;
+ int64_t vmport_guestsize;
+ uint32_t vmport_used_guestinfo;
+ uint32_t vmport_used_guestinfo_jumbo;
+ uint8_t pb[vmport_size];
+ char *p;
+ int chans_size;
+
+ READ(pb);
+ vp = (void *)&pb;
+
+ p = vp->u.packed.packed_data;
+
+ vmport_guestsize = vp->used_guestsize;
+ chans_size = vmport_size - vmport_guestsize - (p - (char *)pb);
+ if ( chans_size < 0 )
+ {
+ fprintf(stderr, "*** VMPORT: bogus chans_size=%d should be >= 0\n"
+ " vmport_size=%d vmport_guestsize=%d fixed_size=%ld\n",
+ chans_size, vmport_size, (int)vmport_guestsize,
+ p - (char *)pb);
+ chans_size = 0;
+ }
+
+ printf(" VMPORT: ping_time %" PRIu64 "\n", vp->ping_time);
+ printf(" VMPORT: open_cookie %" PRIx32 "\n", vp->open_cookie);
+ for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
+ {
+ int j;
+ vmport_channel_control_t *vc = &vp->u.packed.chan_ctl[i].chan;
+ vmport_bucket_control_t *jb = &vp->u.packed.chan_ctl[i].jumbo;
+
+ printf(" VMPORT: chan[%d] chan_id %d\n", i, vc->chan_id);
+ printf(" VMPORT: chan[%d] active_time %" PRIx64 "\n",
+ i, vc->active_time);
+ printf(" VMPORT: chan[%d] proto_num %" PRIx32 "\n",
+ i, vc->proto_num);
+ printf(" VMPORT: chan[%d] recv_read %d\n", i, vc->recv_read);
+ printf(" VMPORT: chan[%d] recv_write %d\n", i, vc->recv_write);
+ printf(" VMPORT: chan[%d] jumbo %d\n", i, vc->jumbo);
+ printf(" VMPORT: chan[%d] send_len %d\n", i, vc->send_len);
+ if ( vc->send_len > VMPORT_MAX_SEND_BUF * 4 )
+ {
+ printf("--- VMPORT: trucated send_len=%d > %d\n",
+ vc->send_len, VMPORT_MAX_SEND_BUF * 4);
+ vc->send_len = VMPORT_MAX_SEND_BUF * 4;
+ }
+ if ( vc->send_len > chans_size )
+ {
+ fprintf(stderr, "*** VMPORT: bogus send_len=%d > %d\n",
+ vc->send_len, chans_size);
+ if ( chans_size >= 0 )
+ vc->send_len = chans_size;
+ else
+ vc->send_len = 0;
+ }
+ p += vc->send_len;
+ chans_size -= vc->send_len;
+ for ( j = 0; j < VMPORT_MAX_BKTS; j++ )
+ {
+ vmport_bucket_control_t *b = &vp->u.packed.chan_ctl[i].recv[j];
+
+ printf(" VMPORT: chan[%d] bucket[%d] recv_len %d\n",
+ i, j, b->recv_len);
+ if ( b->recv_len > VMPORT_MAX_RECV_BUF * 4 )
+ {
+ printf("--- VMPORT: trucated recv_len=%d > %d\n",
+ b->recv_len, VMPORT_MAX_RECV_BUF * 4);
+ b->recv_len = VMPORT_MAX_RECV_BUF * 4;
+ }
+ if ( b->recv_len > chans_size )
+ {
+ fprintf(stderr, "*** VMPORT: bogus recv_len=%d > %d\n",
+ b->recv_len, chans_size);
+ if ( chans_size >= 0 )
+ b->recv_len = chans_size;
+ else
+ b->recv_len = 0;
+ }
+ p += b->recv_len;
+ chans_size -= b->recv_len;
+ }
+ printf(" VMPORT: chan[%d] jumbo_bkt recv_len %d\n", i, jb->recv_len);
+ if ( jb->recv_len > VMPORT_MAX_RECV_JUMBO_BUF * 4 )
+ {
+ printf("--- VMPORT: trucated recv_len=%d > %d\n",
+ jb->recv_len, VMPORT_MAX_RECV_JUMBO_BUF * 4);
+ jb->recv_len = VMPORT_MAX_RECV_JUMBO_BUF * 4;
+ }
+ if ( jb->recv_len > chans_size )
+ {
+ fprintf(stderr, "*** VMPORT: bogus recv_len=%d > %d\n",
+ jb->recv_len, chans_size);
+ if ( chans_size >= 0 )
+ jb->recv_len = chans_size;
+ else
+ jb->recv_len = 0;
+ }
+ p += jb->recv_len;
+ chans_size -= jb->recv_len;
+ }
+
+ if ( chans_size != 0 )
+ fprintf(stderr, "*** VMPORT: bogus chans_size=%d should be 0\n",
+ chans_size);
+
+ vmport_used_guestinfo = vp->used_guestinfo;
+ vmport_used_guestinfo_jumbo = vp->used_guestinfo_jumbo;
+
+ if ( vmport_used_guestinfo == 0 )
+ printf(" VMPORT: no small data\n");
+ for ( i = 0; i < vmport_used_guestinfo; i++ )
+ {
+ if ( vmport_guestsize > 0 )
+ {
+ uint8_t key_len = (uint8_t)(*p++);
+ uint8_t val_len = (uint8_t)(*p++);
+ if ( key_len )
+ {
+ char key[VMPORT_MAX_KEY_LEN + 1];
+ char val[VMPORT_MAX_VAL_LEN + 1];
+
+ if ( key_len > VMPORT_MAX_KEY_LEN )
+ {
+ fprintf(stderr,
+ "*** VMPORT: bogus key_len=%d > %d for guestinfo[%d]\n",
+ key_len, VMPORT_MAX_KEY_LEN, i);
+ key_len = VMPORT_MAX_KEY_LEN;
+ }
+ memcpy(key, p, key_len);
+ p += key_len;
+ key[key_len] = '\0';
+ if ( val_len > VMPORT_MAX_VAL_LEN )
+ {
+ fprintf(stderr,
+ "*** VMPORT: bogus val_len=%d > %d for guestinfo[%d]\n",
+ val_len, VMPORT_MAX_VAL_LEN, i);
+ val_len = VMPORT_MAX_VAL_LEN;
+ }
+ memcpy(val, p, val_len);
+ p += val_len;
+ val[val_len] = '\0';
+ vmport_guestsize -= 2 + key_len + val_len;
+ printf(" VMPORT: guestinfo[%d](%s) = \"%s\"\n",
+ i, key, val);
+ }
+ else
+ {
+ fprintf(stderr,
+ "*** VMPORT: bogus len for guestinfo[%d]\n",
+ i);
+ vmport_guestsize -= 2;
+ }
+ if ( vmport_guestsize < 0 )
+ printf(" VMPORT: data length skew at guestinfo[%d]\n"
+ " remaining datasize=%ld\n",
+ i, vmport_guestsize);
+ }
+ }
+
+ if ( vmport_guestsize == 0 )
+ printf(" VMPORT: no jumbo data\n");
+ for ( i = 0; i < vmport_used_guestinfo_jumbo; i++ )
+ {
+ if ( vmport_guestsize > 0 )
+ {
+ uint8_t key_len = (uint8_t)(*p++);
+ uint16_t val_len;
+
+ memcpy(&val_len, p, 2);
+ p += 2;
+ if ( key_len )
+ {
+ char key[VMPORT_MAX_KEY_LEN + 1];
+ char val[VMPORT_MAX_VAL_JUMBO_LEN + 1];
+
+ if ( key_len > VMPORT_MAX_KEY_LEN )
+ {
+ fprintf(stderr,
+ "*** VMPORT: bogus key_len=%d > %d for guestinfo[%d]\n",
+ key_len, VMPORT_MAX_KEY_LEN, i);
+ key_len = VMPORT_MAX_KEY_LEN;
+ }
+ memcpy(key, p, key_len);
+ p += key_len;
+ key[key_len] = '\0';
+ if ( val_len > VMPORT_MAX_VAL_JUMBO_LEN )
+ {
+ fprintf(stderr,
+ "*** VMPORT: bogus val_len=%d > %d for guestinfo[%d]\n",
+ val_len, VMPORT_MAX_VAL_JUMBO_LEN, i);
+ val_len = VMPORT_MAX_VAL_JUMBO_LEN;
+ }
+ memcpy(val, p, val_len);
+ p += val_len;
+ val[val_len] = '\0';
+ vmport_guestsize -= 2 + key_len + val_len;
+ printf(" VMPORT: guestinfo_jumbo[%d](%s) = \"%s\"\n",
+ i, key, val);
+ }
+ else
+ {
+ printf(" VMPORT: bogus len for guestinfo_jumbo[%d]\n", i);
+ vmport_guestsize -= 2;
+ }
+ if ( vmport_guestsize < 0 )
+ printf(" VMPORT: data length skew at guestinfo_jumbo[%d]\n"
+ " remaining datasize=%ld\n", i,
+ vmport_guestsize);
+ }
+ }
+
+ if ( !vmport_guestsize )
+ printf(" VMPORT: %ld bytes leftover data\n", vmport_guestsize);
+}
+
int main(int argc, char **argv)
{
int entry, domid;
@@ -467,6 +686,7 @@ int main(int argc, char **argv)
case HVM_SAVE_CODE(VIRIDIAN_VCPU): dump_viridian_vcpu(); break;
case HVM_SAVE_CODE(VMCE_VCPU): dump_vmce_vcpu(); break;
case HVM_SAVE_CODE(TSC_ADJUST): dump_tsc_adjust(); break;
+ case HVM_SAVE_CODE(VMPORT): dump_vmport(desc.length); break;
case HVM_SAVE_CODE(END): break;
default:
printf(" ** Don't understand type %u: skipping\n",
@@ -477,3 +697,12 @@ int main(int argc, char **argv)
return 0;
}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 13/16] Add xen-hvm-param
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (11 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 12/16] Add dump of HVM_SAVE_CODE(VMPORT) to xen-hvmctx Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 14/16] Add xen-vmware-guestinfo Don Slutz
` (2 subsequent siblings)
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
A tool to get and set hvm param.
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
.gitignore | 1 +
tools/misc/Makefile | 7 +-
tools/misc/xen-hvm-param.c | 164 +++++++++++++++++++++++++++++++++++++++++++++
3 files changed, 170 insertions(+), 2 deletions(-)
create mode 100644 tools/misc/xen-hvm-param.c
diff --git a/.gitignore b/.gitignore
index bb928a5..aff5758 100644
--- a/.gitignore
+++ b/.gitignore
@@ -180,6 +180,7 @@ tools/misc/xen-tmem-list-parse
tools/misc/xenperf
tools/misc/xenpm
tools/misc/xen-hvmctx
+tools/misc/xen-hvm-param
tools/misc/gtraceview
tools/misc/gtracestat
tools/misc/xenlockprof
diff --git a/tools/misc/Makefile b/tools/misc/Makefile
index 69b1817..b8d4579 100644
--- a/tools/misc/Makefile
+++ b/tools/misc/Makefile
@@ -10,7 +10,7 @@ CFLAGS += $(CFLAGS_libxenstore)
HDRS = $(wildcard *.h)
TARGETS-y := xenperf xenpm xen-tmem-list-parse gtraceview gtracestat xenlockprof xenwatchdogd xencov
-TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvmcrash xen-lowmemd xen-mfndump
+TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvm-param xen-hvmcrash xen-lowmemd xen-mfndump
TARGETS-$(CONFIG_MIGRATE) += xen-hptool
TARGETS := $(TARGETS-y)
@@ -22,7 +22,7 @@ INSTALL_BIN := $(INSTALL_BIN-y)
INSTALL_SBIN-y := xen-bugtool xen-python-path xenperf xenpm xen-tmem-list-parse gtraceview \
gtracestat xenlockprof xenwatchdogd xen-ringwatch xencov
-INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvmcrash xen-lowmemd xen-mfndump
+INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvm-param xen-hvmcrash xen-lowmemd xen-mfndump
INSTALL_SBIN-$(CONFIG_MIGRATE) += xen-hptool
INSTALL_SBIN := $(INSTALL_SBIN-y)
@@ -57,6 +57,9 @@ clean:
xen-hvmctx: xen-hvmctx.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+xen-hvm-param: xen-hvm-param.o
+ $(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+
xen-hvmcrash: xen-hvmcrash.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
diff --git a/tools/misc/xen-hvm-param.c b/tools/misc/xen-hvm-param.c
new file mode 100644
index 0000000..0c62fa2
--- /dev/null
+++ b/tools/misc/xen-hvm-param.c
@@ -0,0 +1,164 @@
+/*
+ * tools/misc/xen-hvm-param.c
+ *
+ * Copyright (C) 2014 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <err.h>
+
+#include <xenctrl.h>
+
+
+int
+main(int argc, char **argv)
+{
+ xc_interface *xch;
+ int domid;
+ int start_param = 0;
+ int end_param = HVM_NR_PARAMS;
+ int param;
+ int ret = 0;
+ int i;
+ char hvm_param_name[HVM_NR_PARAMS][80];
+
+ unsigned long hvm_param = -1;
+
+ if ( (argc < 2) || (argc > 4) )
+ errx(1, "usage: %s domid [param [new]]", argv[0]);
+
+ for ( i = 0; i < HVM_NR_PARAMS; i++ )
+ snprintf(hvm_param_name[i], sizeof(hvm_param_name[i]), "Unknown %d", i);
+
+ snprintf(hvm_param_name[HVM_PARAM_CALLBACK_IRQ],
+ sizeof(hvm_param_name[HVM_PARAM_CALLBACK_IRQ]), "Callback_Irq");
+ snprintf(hvm_param_name[HVM_PARAM_STORE_PFN],
+ sizeof(hvm_param_name[HVM_PARAM_STORE_PFN]), "Store_Pfn");
+ snprintf(hvm_param_name[HVM_PARAM_STORE_EVTCHN],
+ sizeof(hvm_param_name[HVM_PARAM_STORE_EVTCHN]), "Store_Evtchn");
+ snprintf(hvm_param_name[HVM_PARAM_PAE_ENABLED],
+ sizeof(hvm_param_name[HVM_PARAM_PAE_ENABLED]), "Pae_Enabled");
+ snprintf(hvm_param_name[HVM_PARAM_IOREQ_PFN],
+ sizeof(hvm_param_name[HVM_PARAM_IOREQ_PFN]), "Ioreq_Pfn");
+ snprintf(hvm_param_name[HVM_PARAM_BUFIOREQ_PFN],
+ sizeof(hvm_param_name[HVM_PARAM_BUFIOREQ_PFN]), "Bufioreq_Pfn");
+ snprintf(hvm_param_name[HVM_PARAM_VIRIDIAN],
+ sizeof(hvm_param_name[HVM_PARAM_VIRIDIAN]), "Viridian");
+ snprintf(hvm_param_name[HVM_PARAM_TIMER_MODE],
+ sizeof(hvm_param_name[HVM_PARAM_TIMER_MODE]), "Timer_Mode");
+ snprintf(hvm_param_name[HVM_PARAM_HPET_ENABLED],
+ sizeof(hvm_param_name[HVM_PARAM_HPET_ENABLED]), "Hpet_Enabled");
+ snprintf(hvm_param_name[HVM_PARAM_IDENT_PT],
+ sizeof(hvm_param_name[HVM_PARAM_IDENT_PT]), "Ident_Pt");
+ snprintf(hvm_param_name[HVM_PARAM_DM_DOMAIN],
+ sizeof(hvm_param_name[HVM_PARAM_DM_DOMAIN]), "Dm_Domain");
+ snprintf(hvm_param_name[HVM_PARAM_ACPI_S_STATE],
+ sizeof(hvm_param_name[HVM_PARAM_ACPI_S_STATE]), "Acpi_S_State");
+ snprintf(hvm_param_name[HVM_PARAM_VM86_TSS],
+ sizeof(hvm_param_name[HVM_PARAM_VM86_TSS]), "Vm86_Tss");
+ snprintf(hvm_param_name[HVM_PARAM_VPT_ALIGN],
+ sizeof(hvm_param_name[HVM_PARAM_VPT_ALIGN]), "Vpt_Align");
+ snprintf(hvm_param_name[HVM_PARAM_CONSOLE_PFN],
+ sizeof(hvm_param_name[HVM_PARAM_CONSOLE_PFN]), "Console_Pfn");
+ snprintf(hvm_param_name[HVM_PARAM_CONSOLE_EVTCHN],
+ sizeof(hvm_param_name[HVM_PARAM_CONSOLE_EVTCHN]), "Console_Evtchn");
+ snprintf(hvm_param_name[HVM_PARAM_ACPI_IOPORTS_LOCATION],
+ sizeof(hvm_param_name[HVM_PARAM_ACPI_IOPORTS_LOCATION]),
+ "Acpi_Ioports_Location");
+ snprintf(hvm_param_name[HVM_PARAM_MEMORY_EVENT_CR0],
+ sizeof(hvm_param_name[HVM_PARAM_MEMORY_EVENT_CR0]), "Memory_Event_Cr0");
+ snprintf(hvm_param_name[HVM_PARAM_MEMORY_EVENT_CR3],
+ sizeof(hvm_param_name[HVM_PARAM_MEMORY_EVENT_CR3]), "Memory_Event_Cr3");
+ snprintf(hvm_param_name[HVM_PARAM_MEMORY_EVENT_CR4],
+ sizeof(hvm_param_name[HVM_PARAM_MEMORY_EVENT_CR4]), "Memory_Event_Cr4");
+ snprintf(hvm_param_name[HVM_PARAM_MEMORY_EVENT_INT3],
+ sizeof(hvm_param_name[HVM_PARAM_MEMORY_EVENT_INT3]), "Memory_Event_Int3");
+ snprintf(hvm_param_name[HVM_PARAM_NESTEDHVM],
+ sizeof(hvm_param_name[HVM_PARAM_NESTEDHVM]), "Nestedhvm");
+ snprintf(hvm_param_name[HVM_PARAM_MEMORY_EVENT_SINGLE_STEP],
+ sizeof(hvm_param_name[HVM_PARAM_MEMORY_EVENT_SINGLE_STEP]),
+ "Memory_Event_Single_Step");
+ snprintf(hvm_param_name[HVM_PARAM_BUFIOREQ_EVTCHN],
+ sizeof(hvm_param_name[HVM_PARAM_BUFIOREQ_EVTCHN]), "Bufioreq_Evtchn");
+ snprintf(hvm_param_name[HVM_PARAM_PAGING_RING_PFN],
+ sizeof(hvm_param_name[HVM_PARAM_PAGING_RING_PFN]), "Paging_Ring_Pfn");
+ snprintf(hvm_param_name[HVM_PARAM_ACCESS_RING_PFN],
+ sizeof(hvm_param_name[HVM_PARAM_ACCESS_RING_PFN]), "Access_Ring_Pfn");
+ snprintf(hvm_param_name[HVM_PARAM_SHARING_RING_PFN],
+ sizeof(hvm_param_name[HVM_PARAM_SHARING_RING_PFN]), "Sharing_Ring_Pfn");
+ snprintf(hvm_param_name[HVM_PARAM_VMWARE_HW],
+ sizeof(hvm_param_name[HVM_PARAM_VMWARE_HW]), "Vmware_Hw");
+ snprintf(hvm_param_name[HVM_PARAM_VMPORT_BUILD_NUMBER_TIME],
+ sizeof(hvm_param_name[HVM_PARAM_VMPORT_BUILD_NUMBER_TIME]),
+ "Vmport_Build_Number_Time");
+ snprintf(hvm_param_name[HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE],
+ sizeof(hvm_param_name[HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE]),
+ "Vmport_Build_Number_Value");
+ snprintf(hvm_param_name[HVM_PARAM_VMPORT_RESET_TIME],
+ sizeof(hvm_param_name[HVM_PARAM_VMPORT_RESET_TIME]), "Vmport_Reset_Time");
+
+ xch = xc_interface_open(0, 0, 0);
+ if ( !xch )
+ err(1, "failed to open control interface");
+
+ domid = atoi(argv[1]);
+ if ( argc > 2 )
+ {
+ start_param = strtol(argv[2], NULL, 0);
+ end_param = start_param + 1;
+ }
+
+ for ( param = start_param; param < end_param; param++ )
+ {
+ ret = xc_get_hvm_param(xch, domid, param, &hvm_param);
+ if ( ret )
+ err(1, "failed to get hvm param %d for domid %d", param, domid);
+ else
+ {
+ if ( argc == 4 )
+ {
+ long new = strtol(argv[3], NULL, 0);
+
+ ret = xc_set_hvm_param(xch, domid, param, new);
+ if ( ret )
+ err(1, "failed to set hvm param %d for domid %d", param, domid);
+ else if ( (param >= 0) && (param < HVM_NR_PARAMS) )
+ printf("hvm_param(%s)=0x%lx(%ld) was 0x%lx(%ld)\n",
+ hvm_param_name[param], new, new, hvm_param, hvm_param);
+ else
+ printf("hvm_param(%d)=0x%lx(%ld) was 0x%lx(%ld)\n",
+ param, new, new, hvm_param, hvm_param);
+ }
+ else
+ {
+ if ( (param >= 0) && (param < HVM_NR_PARAMS) )
+ printf("hvm_param(%s)=0x%lx(%ld)\n", hvm_param_name[param], hvm_param,
+ hvm_param);
+ else
+ printf("hvm_param(%d)=0x%lx(%ld)\n", param, hvm_param, hvm_param);
+ }
+ }
+ }
+ xc_interface_close(xch);
+
+ return ret;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 14/16] Add xen-vmware-guestinfo
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (12 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 13/16] Add xen-hvm-param Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 15/16] Add xen-list-vmware-guestinfo Don Slutz
2014-09-11 18:36 ` [PATCH v4 16/16] Add xen-hvm-send-trigger Don Slutz
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
A tool to get and set VMware guestinfo
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
.gitignore | 1 +
tools/misc/Makefile | 7 ++-
tools/misc/xen-vmware-guestinfo.c | 97 +++++++++++++++++++++++++++++++++++++++
3 files changed, 103 insertions(+), 2 deletions(-)
create mode 100644 tools/misc/xen-vmware-guestinfo.c
diff --git a/.gitignore b/.gitignore
index aff5758..07f20b9 100644
--- a/.gitignore
+++ b/.gitignore
@@ -181,6 +181,7 @@ tools/misc/xenperf
tools/misc/xenpm
tools/misc/xen-hvmctx
tools/misc/xen-hvm-param
+tools/misc/xen-vmware-guestinfo
tools/misc/gtraceview
tools/misc/gtracestat
tools/misc/xenlockprof
diff --git a/tools/misc/Makefile b/tools/misc/Makefile
index b8d4579..f2ffe1a 100644
--- a/tools/misc/Makefile
+++ b/tools/misc/Makefile
@@ -10,7 +10,7 @@ CFLAGS += $(CFLAGS_libxenstore)
HDRS = $(wildcard *.h)
TARGETS-y := xenperf xenpm xen-tmem-list-parse gtraceview gtracestat xenlockprof xenwatchdogd xencov
-TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvm-param xen-hvmcrash xen-lowmemd xen-mfndump
+TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
TARGETS-$(CONFIG_MIGRATE) += xen-hptool
TARGETS := $(TARGETS-y)
@@ -22,7 +22,7 @@ INSTALL_BIN := $(INSTALL_BIN-y)
INSTALL_SBIN-y := xen-bugtool xen-python-path xenperf xenpm xen-tmem-list-parse gtraceview \
gtracestat xenlockprof xenwatchdogd xen-ringwatch xencov
-INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvm-param xen-hvmcrash xen-lowmemd xen-mfndump
+INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
INSTALL_SBIN-$(CONFIG_MIGRATE) += xen-hptool
INSTALL_SBIN := $(INSTALL_SBIN-y)
@@ -60,6 +60,9 @@ xen-hvmctx: xen-hvmctx.o
xen-hvm-param: xen-hvm-param.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+xen-vmware-guestinfo: xen-vmware-guestinfo.o
+ $(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+
xen-hvmcrash: xen-hvmcrash.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
diff --git a/tools/misc/xen-vmware-guestinfo.c b/tools/misc/xen-vmware-guestinfo.c
new file mode 100644
index 0000000..e6b288c
--- /dev/null
+++ b/tools/misc/xen-vmware-guestinfo.c
@@ -0,0 +1,97 @@
+/*
+ * tools/misc/xen-vmware-guestinfo.c
+ *
+ * Copyright (C) 2014 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <err.h>
+
+#include <xenctrl.h>
+
+
+int
+main(int argc, char **argv)
+{
+ xc_interface *xch;
+ int domid;
+ int ret = 0;
+
+ char value[8192];
+ unsigned int val_len;
+ char *vals = "";
+
+ if ( (argc < 3) || (argc > 4) )
+ errx(1, "usage: %s domid guestinfo.<name> [new]", argv[0]);
+
+ xch = xc_interface_open(0, 0, 0);
+ if ( !xch )
+ err(1, "failed to open control interface");
+
+ domid = atoi(argv[1]);
+
+ ret = xc_get_vmport_guest_info(xch, domid, strlen(argv[2]), argv[2],
+ sizeof(value), &val_len, value);
+ if ( !ret )
+ {
+ /* Make sure this is a c-string */
+ if ( val_len < sizeof(value) )
+ value[val_len] = 0;
+ else
+ {
+ value[sizeof(value) - 1] = 0;
+ vals = "...";
+ }
+ }
+
+ if ( argc == 4 )
+ {
+ int ret1;
+
+ if ( ret )
+ warn("failed to get VMware guestinfo '%s' for domid %d", argv[2], domid);
+ ret1 = xc_set_vmport_guest_info(xch, domid, strlen(argv[2]), argv[2],
+ strlen(argv[3]), argv[3]);
+ if ( ret1 )
+ err(1, "failed to set VMware guestinfo '%s' for domid %d", argv[2], domid);
+ else if ( ret )
+ printf("VMware guestinfo '%s'='%s'\n",
+ argv[2], argv[3]);
+ else
+ printf("VMware guestinfo '%s' was '%s'%s now '%s'\n",
+ argv[2], value, vals, argv[3]);
+ }
+ else
+ {
+ if ( ret )
+ err(1, "failed to get VMware guestinfo '%s' for domid %d", argv[2], domid);
+ else
+ {
+ printf("VMware guestinfo '%s'='%s'%s\n",
+ argv[2], value, vals);
+ }
+ }
+ xc_interface_close(xch);
+
+ return ret;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 15/16] Add xen-list-vmware-guestinfo
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (13 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 14/16] Add xen-vmware-guestinfo Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 16/16] Add xen-hvm-send-trigger Don Slutz
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
A tool to list currently set VMware guestinfo
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
.gitignore | 1 +
tools/misc/Makefile | 7 ++-
tools/misc/xen-list-vmware-guestinfo.c | 88 ++++++++++++++++++++++++++++++++++
3 files changed, 94 insertions(+), 2 deletions(-)
create mode 100644 tools/misc/xen-list-vmware-guestinfo.c
diff --git a/.gitignore b/.gitignore
index 07f20b9..606c703 100644
--- a/.gitignore
+++ b/.gitignore
@@ -182,6 +182,7 @@ tools/misc/xenpm
tools/misc/xen-hvmctx
tools/misc/xen-hvm-param
tools/misc/xen-vmware-guestinfo
+tools/misc/xen-list-vmware-guestinfo
tools/misc/gtraceview
tools/misc/gtracestat
tools/misc/xenlockprof
diff --git a/tools/misc/Makefile b/tools/misc/Makefile
index f2ffe1a..3e7d216 100644
--- a/tools/misc/Makefile
+++ b/tools/misc/Makefile
@@ -10,7 +10,7 @@ CFLAGS += $(CFLAGS_libxenstore)
HDRS = $(wildcard *.h)
TARGETS-y := xenperf xenpm xen-tmem-list-parse gtraceview gtracestat xenlockprof xenwatchdogd xencov
-TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
+TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-list-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
TARGETS-$(CONFIG_MIGRATE) += xen-hptool
TARGETS := $(TARGETS-y)
@@ -22,7 +22,7 @@ INSTALL_BIN := $(INSTALL_BIN-y)
INSTALL_SBIN-y := xen-bugtool xen-python-path xenperf xenpm xen-tmem-list-parse gtraceview \
gtracestat xenlockprof xenwatchdogd xen-ringwatch xencov
-INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
+INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-list-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
INSTALL_SBIN-$(CONFIG_MIGRATE) += xen-hptool
INSTALL_SBIN := $(INSTALL_SBIN-y)
@@ -63,6 +63,9 @@ xen-hvm-param: xen-hvm-param.o
xen-vmware-guestinfo: xen-vmware-guestinfo.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+xen-list-vmware-guestinfo: xen-list-vmware-guestinfo.o
+ $(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+
xen-hvmcrash: xen-hvmcrash.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
diff --git a/tools/misc/xen-list-vmware-guestinfo.c b/tools/misc/xen-list-vmware-guestinfo.c
new file mode 100644
index 0000000..2d253b6
--- /dev/null
+++ b/tools/misc/xen-list-vmware-guestinfo.c
@@ -0,0 +1,88 @@
+/*
+ * tools/misc/xen-list-vmware-guestinfo.c
+ *
+ * Copyright (C) 2014 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <err.h>
+
+#include <xenctrl.h>
+
+
+int
+main(int argc, char **argv)
+{
+ xc_interface *xch;
+ int domid;
+ int ret = 0;
+
+ unsigned int idx = 0;
+ char key[128];
+ unsigned int key_len;
+ char value[8192];
+ unsigned int value_len;
+
+ if ( argc != 2 )
+ errx(1, "usage: %s domid", argv[0]);
+
+ xch = xc_interface_open(0, 0, 0);
+ if ( !xch )
+ err(1, "failed to open control interface");
+
+ domid = atoi(argv[1]);
+
+ while ( !xc_fetch_vmport_guest_info(xch, domid, idx, sizeof(key),
+ &key_len, key, sizeof(value),
+ &value_len, value) )
+ {
+ if ( key_len )
+ {
+ char *keys = "";
+ char *vals = "";
+
+ /* Make sure this is a c-string */
+ if ( key_len < sizeof(key) )
+ key[key_len] = 0;
+ else
+ {
+ key[sizeof(key) - 1] = 0;
+ keys = "...";
+ }
+ /* Make sure this is a c-string */
+ if ( value_len < sizeof(value) )
+ value[value_len] = 0;
+ else
+ {
+ value[sizeof(value) - 1] = 0;
+ vals = "...";
+ }
+ printf("VMware guestinfo '%s'%s='%s'%s\n",
+ key, keys, value, vals);
+ }
+ idx++;
+ }
+ xc_interface_close(xch);
+
+ return ret;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* [PATCH v4 16/16] Add xen-hvm-send-trigger
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
` (14 preceding siblings ...)
2014-09-11 18:36 ` [PATCH v4 15/16] Add xen-list-vmware-guestinfo Don Slutz
@ 2014-09-11 18:36 ` Don Slutz
15 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-11 18:36 UTC (permalink / raw)
To: xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Don Slutz, Tim Deegan,
George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Boris Ostrovsky, Suravee Suthikulpanit
A tool to send a trigger to a domU via xc_domain_send_trigger
Signed-off-by: Don Slutz <dslutz@verizon.com>
---
.gitignore | 1 +
tools/misc/Makefile | 7 ++-
tools/misc/xen-hvm-send-trigger.c | 103 ++++++++++++++++++++++++++++++++++++++
3 files changed, 109 insertions(+), 2 deletions(-)
create mode 100644 tools/misc/xen-hvm-send-trigger.c
diff --git a/.gitignore b/.gitignore
index 606c703..d66c5f5 100644
--- a/.gitignore
+++ b/.gitignore
@@ -183,6 +183,7 @@ tools/misc/xen-hvmctx
tools/misc/xen-hvm-param
tools/misc/xen-vmware-guestinfo
tools/misc/xen-list-vmware-guestinfo
+tools/misc/xen-hvm-send-trigger
tools/misc/gtraceview
tools/misc/gtracestat
tools/misc/xenlockprof
diff --git a/tools/misc/Makefile b/tools/misc/Makefile
index 3e7d216..9c5e988 100644
--- a/tools/misc/Makefile
+++ b/tools/misc/Makefile
@@ -10,7 +10,7 @@ CFLAGS += $(CFLAGS_libxenstore)
HDRS = $(wildcard *.h)
TARGETS-y := xenperf xenpm xen-tmem-list-parse gtraceview gtracestat xenlockprof xenwatchdogd xencov
-TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-list-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
+TARGETS-$(CONFIG_X86) += xen-detect xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-list-vmware-guestinfo xen-hvm-send-trigger xen-hvmcrash xen-lowmemd xen-mfndump
TARGETS-$(CONFIG_MIGRATE) += xen-hptool
TARGETS := $(TARGETS-y)
@@ -22,7 +22,7 @@ INSTALL_BIN := $(INSTALL_BIN-y)
INSTALL_SBIN-y := xen-bugtool xen-python-path xenperf xenpm xen-tmem-list-parse gtraceview \
gtracestat xenlockprof xenwatchdogd xen-ringwatch xencov
-INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-list-vmware-guestinfo xen-hvmcrash xen-lowmemd xen-mfndump
+INSTALL_SBIN-$(CONFIG_X86) += xen-hvmctx xen-hvm-param xen-vmware-guestinfo xen-list-vmware-guestinfo xen-hvm-send-trigger xen-hvmcrash xen-lowmemd xen-mfndump
INSTALL_SBIN-$(CONFIG_MIGRATE) += xen-hptool
INSTALL_SBIN := $(INSTALL_SBIN-y)
@@ -66,6 +66,9 @@ xen-vmware-guestinfo: xen-vmware-guestinfo.o
xen-list-vmware-guestinfo: xen-list-vmware-guestinfo.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+xen-hvm-send-trigger: xen-hvm-send-trigger.o
+ $(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
+
xen-hvmcrash: xen-hvmcrash.o
$(CC) $(LDFLAGS) -o $@ $< $(LDLIBS_libxenctrl) $(APPEND_LDFLAGS)
diff --git a/tools/misc/xen-hvm-send-trigger.c b/tools/misc/xen-hvm-send-trigger.c
new file mode 100644
index 0000000..b822f9c
--- /dev/null
+++ b/tools/misc/xen-hvm-send-trigger.c
@@ -0,0 +1,103 @@
+/*
+ * tools/misc/xen-hvm-send-trigger.c
+ *
+ * Copyright (C) 2014 Verizon Corporation
+ *
+ * This file is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License Version 2 (GPLv2)
+ * as published by the Free Software Foundation.
+ *
+ * This file is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * General Public License for more details. <http://www.gnu.org/licenses/>.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <err.h>
+
+#include <xenctrl.h>
+#include <xen/domctl.h>
+
+#define HVM_NR_SENDTRIGGERS 8
+
+int
+main(int argc, char **argv)
+{
+ xc_interface *xch;
+ int domid;
+ int sendtrigger;
+ uint32_t vcpuid = 0;
+ int ret = 0;
+ int i;
+ char hvm_sendtrigger_name[HVM_NR_SENDTRIGGERS][80];
+
+ if ( (argc < 3) || (argc > 4) )
+ errx(1, "usage: %s domid trigger [vcpuid]", argv[0]);
+
+ for ( i = 0; i < HVM_NR_SENDTRIGGERS; i++ )
+ snprintf(hvm_sendtrigger_name[i], sizeof(hvm_sendtrigger_name[i]), "Unknown %d",
+ i);
+
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_NMI],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_NMI]),
+ "Trigger_Nmi");
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_RESET],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_RESET]),
+ "Trigger_Reset");
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_INIT],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_INIT]),
+ "Trigger_Init");
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_POWER],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_POWER]),
+ "Trigger_Power");
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_SLEEP],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_SLEEP]),
+ "Trigger_Sleep");
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_VTPOWER],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_VTPOWER]),
+ "Trigger_VTPower");
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_VTREBOOT],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_VTREBOOT]),
+ "Trigger_VTReboot");
+ snprintf(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_VTPING],
+ sizeof(hvm_sendtrigger_name[XEN_DOMCTL_SENDTRIGGER_VTPING]),
+ "Trigger_VTPing");
+
+
+ xch = xc_interface_open(0, 0, 0);
+ if ( !xch )
+ err(1, "failed to open control interface");
+
+ domid = atoi(argv[1]);
+ sendtrigger = strtol(argv[2], NULL, 0);
+ if ( argc > 3 )
+ vcpuid = strtol(argv[3], NULL, 0);
+
+ if ( sendtrigger >= 0 && sendtrigger < HVM_NR_SENDTRIGGERS )
+ printf("Sending trigger(%s)=%d to domid %d vcpuid %u\n",
+ hvm_sendtrigger_name[sendtrigger], sendtrigger,
+ domid, vcpuid);
+ else
+ printf("Sending trigger %d to domid %d vcpuid %u\n",
+ sendtrigger, domid, vcpuid);
+
+ ret = xc_domain_send_trigger(xch, domid, sendtrigger, vcpuid);
+ if ( ret )
+ err(1, "failed to send sendtrigger %d for domid %d", sendtrigger,
+ domid);
+
+ xc_interface_close(xch);
+
+ return ret;
+}
+
+/*
+ * Local variables:
+ * mode: C
+ * c-file-style: "BSD"
+ * c-basic-offset: 4
+ * indent-tabs-mode: nil
+ * End:
+ */
--
1.8.4
^ permalink raw reply related [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-11 18:36 ` [PATCH v4 01/16] xen: Add support for VMware cpuid leaves Don Slutz
@ 2014-09-11 19:49 ` Andrew Cooper
2014-09-12 9:49 ` Jan Beulich
2014-09-12 21:26 ` Don Slutz
2014-09-12 12:37 ` Boris Ostrovsky
1 sibling, 2 replies; 51+ messages in thread
From: Andrew Cooper @ 2014-09-11 19:49 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Boris Ostrovsky,
Suravee Suthikulpanit
On 11/09/2014 19:36, Don Slutz wrote:
> This is done by adding HVM_PARAM_VMWARE_HW. It is set to the VMware
> virtual hardware version.
>
> Currently 0, 3-4, 6-11 are good values. However the
> code only checks for == 0 or != 0.
>
> If non-zero then
> Return VMware's cpuid leaves.
>
> The support of hypervisor cpuid leaves has not been agreed to.
>
> MicroSoft Hyper-V (AKA viridian) currently must be at 0x40000000.
>
> VMware currently must be at 0x40000000.
>
> KVM currently must be at 0x40000000 (from Seabios).
>
> Seabios will find xen at 0x40000000, 0x40000100, 0x40000200 ..
> 0x40010000.
>
> http://download.microsoft.com/download/F/B/0/FB0D01A3-8E3A-4F5F-AA59-08C8026D3B8A/requirements-for-implementing-microsoft-hypervisor-interface.docx
>
> http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009458
>
> http://lwn.net/Articles/301888/
> Attempted to get this cleaned up.
>
> So based on this, I picked the order:
>
> Xen at 0x40000000 or
> Viridian or VMware at 0x40000000 and Xen at 0x40000100
>
> If both Viridian and VMware selected, report an error.
>
> Signed-off-by: Don Slutz <dslutz@verizon.com>
> ---
> xen/arch/x86/hvm/Makefile | 3 +-
> xen/arch/x86/hvm/hvm.c | 32 +++++++++++++++++++
> xen/arch/x86/hvm/vmware/Makefile | 1 +
> xen/arch/x86/hvm/vmware/cpuid.c | 69 ++++++++++++++++++++++++++++++++++++++++
> xen/arch/x86/traps.c | 8 +++--
> xen/include/asm-x86/hvm/hvm.h | 3 ++
> xen/include/asm-x86/hvm/vmware.h | 31 ++++++++++++++++++
> xen/include/public/hvm/params.h | 5 ++-
> 8 files changed, 148 insertions(+), 4 deletions(-)
> create mode 100644 xen/arch/x86/hvm/vmware/Makefile
> create mode 100644 xen/arch/x86/hvm/vmware/cpuid.c
> create mode 100644 xen/include/asm-x86/hvm/vmware.h
>
> diff --git a/xen/arch/x86/hvm/Makefile b/xen/arch/x86/hvm/Makefile
> index eea5555..77598a6 100644
> --- a/xen/arch/x86/hvm/Makefile
> +++ b/xen/arch/x86/hvm/Makefile
> @@ -1,5 +1,6 @@
> subdir-y += svm
> subdir-y += vmx
> +subdir-y += vmware
>
> obj-y += asid.o
> obj-y += emulate.o
> @@ -22,4 +23,4 @@ obj-y += vlapic.o
> obj-y += vmsi.o
> obj-y += vpic.o
> obj-y += vpt.o
> -obj-y += vpmu.o
> \ No newline at end of file
> +obj-y += vpmu.o
This hunk is unrelated, but is perhaps something better fixed. A
passing note in the commit message perhaps?
> diff --git a/xen/arch/x86/hvm/hvm.c b/xen/arch/x86/hvm/hvm.c
> index 8d905d3..03a1a19 100644
> --- a/xen/arch/x86/hvm/hvm.c
> +++ b/xen/arch/x86/hvm/hvm.c
> @@ -57,6 +57,7 @@
> #include <asm/hvm/cacheattr.h>
> #include <asm/hvm/trace.h>
> #include <asm/hvm/nestedhvm.h>
> +#include <asm/hvm/vmware.h>
> #include <asm/mtrr.h>
> #include <asm/apic.h>
> #include <public/sched.h>
> @@ -4228,6 +4229,9 @@ void hvm_cpuid(unsigned int input, unsigned int *eax, unsigned int *ebx,
> if ( cpuid_viridian_leaves(input, eax, ebx, ecx, edx) )
> return;
>
> + if ( cpuid_vmware_leaves(input, eax, ebx, ecx, edx) )
> + return;
> +
> if ( cpuid_hypervisor_leaves(input, count, eax, ebx, ecx, edx) )
> return;
>
> @@ -5555,6 +5559,11 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
> rc = -EINVAL;
> break;
> case HVM_PARAM_VIRIDIAN:
> + if ( d->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW] )
> + {
> + rc = -EXDEV;
> + break;
> + }
> if ( a.value > 1 )
> rc = -EINVAL;
> break;
> @@ -5692,6 +5701,29 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
>
> break;
> }
> + case HVM_PARAM_VMWARE_HW:
> + /*
> + * This should only ever be set non-zero one time by
> + * the tools and is read only by the guest.
> + */
> + if ( d == current->domain )
> + {
> + rc = -EPERM;
> + break;
> + }
> + if ( d->arch.hvm_domain.params[HVM_PARAM_VIRIDIAN] )
> + {
> + rc = -EXDEV;
> + break;
> + }
> + if ( d->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW] &&
> + d->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW] !=
> + a.value )
> + {
> + rc = -EEXIST;
> + break;
> + }
> + break;
Eugh - its code like this which is why I am certain that things such as
VIRIDIAN, VMWARE (and NESTED_VIRT for that matter) should be domain
creation flags rather than hvm params, and completely immutable for the
entire lifetime of the domain. After all, these params are only being
used as glorified, set-once booleans.
I even have an in-progress patch to turn VIRIDIAN into a create flag
(which has now been complicated by VIRIDIAN turning into a bitmask), but
it appear that doing this necessitates breaking the Libxl API. For
reasons which elude me, the Libxl API exposes the internal
implementation details between "domain creation information" and "domain
build information" in such a way that I couldn't find a compatible way
of moving information from from the latter to the former. (i.e. you
can't duplicate the same field in the create info and still have a user
of an old libxl API version still have their domain creation work when
they only specify the value in the build info)
I am not sure how "fixing things correctly in Xen" fairs against "libxl
taking pain and possibly an API breakage because it previously exposed
internal details which it shouldn't have done", but I would prefer that
we didn't make the problem any harder to fix than it already is.
As a result, I am formally suggesting that this would be better done by
adding a domain creation flag (although not being a maintainer, I
realise my views in this matter don't strictly count for much).
> }
>
> if ( rc == 0 )
> diff --git a/xen/arch/x86/hvm/vmware/Makefile b/xen/arch/x86/hvm/vmware/Makefile
> new file mode 100644
> index 0000000..3fb2e0b
> --- /dev/null
> +++ b/xen/arch/x86/hvm/vmware/Makefile
> @@ -0,0 +1 @@
> +obj-y += cpuid.o
> diff --git a/xen/arch/x86/hvm/vmware/cpuid.c b/xen/arch/x86/hvm/vmware/cpuid.c
> new file mode 100644
> index 0000000..730ab8f
> --- /dev/null
> +++ b/xen/arch/x86/hvm/vmware/cpuid.c
> @@ -0,0 +1,69 @@
> +/*
> + * arch/x86/hvm/vmware/cpuid.c
> + *
> + * Copyright (C) 2012 Verizon Corporation
> + *
> + * This file is free software; you can redistribute it and/or modify it
> + * under the terms of the GNU General Public License Version 2 (GPLv2)
> + * as published by the Free Software Foundation.
> + *
> + * This file is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + * General Public License for more details. <http://www.gnu.org/licenses/>.
> + */
> +
> +#include <xen/sched.h>
> +
> +#include <asm/hvm/hvm.h>
> +#include <asm/hvm/vmware.h>
> +
> +int cpuid_vmware_leaves(uint32_t idx, uint32_t *eax, uint32_t *ebx,
> + uint32_t *ecx, uint32_t *edx)
> +{
> + struct domain *d = current->domain;
> +
> + if ( !is_vmware_domain(d) )
> + return 0;
> +
> + switch ( idx - 0x40000000 )
> + {
> + case 0x0:
Do these leaves have semantic names from a spec, which could live in an
appropriate header file?
> + *eax = 0x40000010; /* Largest leaf */
> + *ebx = 0x61774d56; /* "VMwa" */
> + *ecx = 0x4d566572; /* "reVM" */
> + *edx = 0x65726177; /* "ware" */
> + break;
> +
> + case 0x1 ... 0xf:
> + *eax = 0; /* Reserved */
> + *ebx = 0; /* Reserved */
> + *ecx = 0; /* Reserved */
> + *edx = 0; /* Reserved */
> + break;
> +
> + case 0x10:
> + /* (Virtual) TSC frequency in kHz. */
> + *eax = d->arch.tsc_khz;
> + /* (Virtual) Bus (local apic timer) frequency in kHz. */
> + *ebx = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
At least 1 pair of brackets please, especially as the placement of
brackets affects the result of this particular calculation.
> + *ecx = 0; /* Reserved */
> + *edx = 0; /* Reserved */
> + break;
> +
> + default:
> + return 0;
> + }
> +
> + return 1;
> +}
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * tab-width: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
> index 10fc2ca..f353f42 100644
> --- a/xen/arch/x86/traps.c
> +++ b/xen/arch/x86/traps.c
> @@ -685,8 +685,12 @@ int cpuid_hypervisor_leaves( uint32_t idx, uint32_t sub_idx,
> uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx)
> {
> struct domain *d = current->domain;
> - /* Optionally shift out of the way of Viridian architectural leaves. */
> - uint32_t base = is_viridian_domain(d) ? 0x40000100 : 0x40000000;
> + /*
> + * Optionally shift out of the way of Viridian or VMware
> + * architectural leaves.
> + */
> + uint32_t base = is_viridian_domain(d) | is_vmware_domain(d) ?
> + 0x40000100 : 0x40000000;
Again, brackets please for binary operators. (We have had one recent
slipup because of the precedence of | and ?:)
Furthermore, I think you mean (is_viridian_domain(d) ||
is_vmware_domain(d)) ? 0x40000100 : 0x40000000; which allows for short
circuiting of is_vmware_domain().
> uint32_t limit, dummy;
>
> idx -= base;
> diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
> index 1123857..546210a 100644
> --- a/xen/include/asm-x86/hvm/hvm.h
> +++ b/xen/include/asm-x86/hvm/hvm.h
> @@ -347,6 +347,9 @@ static inline unsigned long hvm_get_shadow_gs_base(struct vcpu *v)
> #define is_viridian_domain(_d) \
> (is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VIRIDIAN]))
>
> +#define is_vmware_domain(_d) \
> + (is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW]))
> +
> void hvm_hypervisor_cpuid_leaf(uint32_t sub_idx,
> uint32_t *eax, uint32_t *ebx,
> uint32_t *ecx, uint32_t *edx);
> diff --git a/xen/include/asm-x86/hvm/vmware.h b/xen/include/asm-x86/hvm/vmware.h
> new file mode 100644
> index 0000000..f254106
> --- /dev/null
> +++ b/xen/include/asm-x86/hvm/vmware.h
> @@ -0,0 +1,31 @@
> +/*
> + * asm-x86/hvm/vmware.h
> + *
> + * Copyright (C) 2012 Verizon Corporation
> + *
> + * This file is free software; you can redistribute it and/or modify it
> + * under the terms of the GNU General Public License Version 2 (GPLv2)
> + * as published by the Free Software Foundation.
> + *
> + * This file is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + * General Public License for more details. <http://www.gnu.org/licenses/>.
> + */
> +
> +#ifndef ASM_X86_HVM_VMWARE_H__
> +#define ASM_X86_HVM_VMWARE_H__
#include <xen/types.h> (IIRC) please, or uint32_t could cause a
compilation failure given specific orderings of #include's in
translation units.
~Andrew
> +
> +int cpuid_vmware_leaves(uint32_t idx, uint32_t *eax, uint32_t *ebx,
> + uint32_t *ecx, uint32_t *edx);
> +
> +#endif /* ASM_X86_HVM_VMWARE_H__ */
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/include/public/hvm/params.h b/xen/include/public/hvm/params.h
> index 614ff5f..dee6d68 100644
> --- a/xen/include/public/hvm/params.h
> +++ b/xen/include/public/hvm/params.h
> @@ -151,6 +151,9 @@
> /* Location of the VM Generation ID in guest physical address space. */
> #define HVM_PARAM_VM_GENERATION_ID_ADDR 34
>
> -#define HVM_NR_PARAMS 35
> +/* Params for VMware */
> +#define HVM_PARAM_VMWARE_HW 35
> +
> +#define HVM_NR_PARAMS 36
>
> #endif /* __XEN_PUBLIC_HVM_PARAMS_H__ */
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 02/16] tools: Add vmware_hw support
2014-09-11 18:36 ` [PATCH v4 02/16] tools: Add vmware_hw support Don Slutz
@ 2014-09-11 21:09 ` Andrew Cooper
2014-09-16 16:20 ` Don Slutz
0 siblings, 1 reply; 51+ messages in thread
From: Andrew Cooper @ 2014-09-11 21:09 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
George Dunlap, Ian Jackson, Eddie Dong, Tim Deegan, Jan Beulich,
Aravind Gopalakrishnan, Jun Nakajima, Boris Ostrovsky,
Suravee Suthikulpanit
On 11/09/2014 19:36, Don Slutz wrote:
> This is used to set HVM_PARAM_VMWARE_HW. It is set to the VMware
> virtual hardware version.
>
> Currently 0, 3-4, 6-11 are good values. However the code only
> checks for == 0 or != 0.
>
> If non-zero then
> default VGA to VMware's VGA.
>
> Also now allows vga=vmware
>
> Signed-off-by: Don Slutz <dslutz@verizon.com>
> ---
> docs/man/xl.cfg.pod.5 | 21 +++++++++++++++++++--
> docs/misc/hypervisor-cpuid.markdown | 29 +++++++++++++++++++++++++++++
> tools/libxc/xc_domain_restore.c | 14 ++++++++++++++
> tools/libxc/xc_domain_save.c | 11 +++++++++++
> tools/libxc/xg_save_restore.h | 2 ++
> tools/libxl/libxl.h | 10 ++++++++++
> tools/libxl/libxl_create.c | 4 +++-
> tools/libxl/libxl_dm.c | 10 +++++++++-
> tools/libxl/libxl_dom.c | 2 ++
> tools/libxl/libxl_types.idl | 2 ++
> tools/libxl/xl_cmdimpl.c | 11 ++++++++++-
> 11 files changed, 111 insertions(+), 5 deletions(-)
> create mode 100644 docs/misc/hypervisor-cpuid.markdown
>
> diff --git a/docs/man/xl.cfg.pod.5 b/docs/man/xl.cfg.pod.5
> index 517ae2f..367b401 100644
> --- a/docs/man/xl.cfg.pod.5
> +++ b/docs/man/xl.cfg.pod.5
> @@ -1147,6 +1147,23 @@ some other Operating Systems and in some circumstance can prevent
> Xen's own paravirtualisation interfaces for HVM guests from being
> used.
>
> +=item B<vmware_hw=NUMBER>
> +
> +Turns on or off the exposure of VMware cpuid. The number is the
> +VMware's hardware version number, where 0 is off. If not zero it
> +changes the default VGA to VMware's VGA.
> +
> +The hardware version number (vmware_hw) come from VMware config files.
> +
> +=over 4
> +
> +In a .vmx it is virtualHW.version
> +
> +In a .ovf it is part of the value of vssd:VirtualSystemType.
> +For vssd:VirtualSystemType == vmx-07, vmware_hw = 7.
> +
> +=back
> +
> =back
>
> =head3 Emulated VGA Graphics Device
> @@ -1185,8 +1202,8 @@ This option is deprecated, use vga="stdvga" instead.
>
> =item B<vga="STRING">
>
> -Selects the emulated video card (none|stdvga|cirrus).
> -The default is cirrus.
> +Selects the emulated video card (none|stdvga|cirrus|vmware).
> +The default is cirrus (or vmware if B<vmware_hw> is not zero).
>
> =item B<vnc=BOOLEAN>
>
> diff --git a/docs/misc/hypervisor-cpuid.markdown b/docs/misc/hypervisor-cpuid.markdown
> new file mode 100644
> index 0000000..4199d57
> --- /dev/null
> +++ b/docs/misc/hypervisor-cpuid.markdown
> @@ -0,0 +1,29 @@
> +Hypervisor Cpuid
> +================
> +
> +The support of hypervisor cpuid leaves has not been agreed to.
> +Other then the range 0x40000000 to 0x400000ff can be used by
> +hypervisors.
> +
> +MicroSoft Hyper-V (AKA viridian) currently must be at 0x40000000.
> +
> +VMware currently must be at 0x40000000.
> +
> +KVM currently must be at 0x40000000 (from Seabios).
> +
> +Seabios will find Xen at 0x40000000, 0x40000100, 0x40000200 ..
> +0x40010000.
Anything looking for Xen according to the Xen cpuid instructions in the
public header files will find Xen in this manner; hvmloader and
HVMLinux to name but a few. I don't think Seabios wants calling out
specifically.
Perhaps "Xen can be found at the first otherwise unused 0x100 aligned
offset between 0x40000000 and 0x40010000"?
~Andrew
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-11 18:36 ` [PATCH v4 04/16] xen: Add is_vmware_port_enabled Don Slutz
@ 2014-09-11 21:22 ` Andrew Cooper
2014-09-16 16:20 ` Don Slutz
2014-09-12 13:08 ` Boris Ostrovsky
1 sibling, 1 reply; 51+ messages in thread
From: Andrew Cooper @ 2014-09-11 21:22 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Boris Ostrovsky,
Suravee Suthikulpanit
On 11/09/2014 19:36, Don Slutz wrote:
> --- a/xen/arch/x86/hvm/vmx/vmx.c
> +++ b/xen/arch/x86/hvm/vmx/vmx.c
> @@ -44,6 +44,7 @@
> #include <asm/hvm/support.h>
> #include <asm/hvm/vmx/vmx.h>
> #include <asm/hvm/vmx/vmcs.h>
> +#include <asm/hvm/vmport.h>
> #include <public/sched.h>
> #include <public/hvm/ioreq.h>
> #include <asm/hvm/vpic.h>
> @@ -1279,6 +1280,8 @@ static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr)
> v->arch.hvm_vmx.exception_bitmap = HVM_TRAP_MASK
> | (paging_mode_hap(v->domain) ?
> 0 : (1U << TRAP_page_fault))
> + | v->domain->arch.hvm_domain.is_vmware_port_enabled ?
> + (1U << TRAP_gp_fault) : 0
I am fairly certain that you need some brackets here.
~Andrew
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-11 19:49 ` Andrew Cooper
@ 2014-09-12 9:49 ` Jan Beulich
2014-09-12 17:46 ` Don Slutz
2014-09-12 21:26 ` Don Slutz
1 sibling, 1 reply; 51+ messages in thread
From: Jan Beulich @ 2014-09-12 9:49 UTC (permalink / raw)
To: Andrew Cooper, xen-devel, Don Slutz
Cc: Jun Nakajima, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Eddie Dong, Ian Jackson,
Tim Deegan, Aravind Gopalakrishnan, Suravee Suthikulpanit,
Boris Ostrovsky
>>> On 11.09.14 at 21:49, <andrew.cooper3@citrix.com> wrote:
> I am not sure how "fixing things correctly in Xen" fairs against "libxl
> taking pain and possibly an API breakage because it previously exposed
> internal details which it shouldn't have done", but I would prefer that
> we didn't make the problem any harder to fix than it already is.
>
> As a result, I am formally suggesting that this would be better done by
> adding a domain creation flag (although not being a maintainer, I
> realise my views in this matter don't strictly count for much).
The main reservation I have against this is that this, in the long run,
may result in a proliferation of VM creation flags. The slightly abused
HVM params at least make adding support of another kind of foreign
VMM emulation a pretty straightforward and isolated change.
>> + case 0x10:
>> + /* (Virtual) TSC frequency in kHz. */
>> + *eax = d->arch.tsc_khz;
>> + /* (Virtual) Bus (local apic timer) frequency in kHz. */
>> + *ebx = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
>
> At least 1 pair of brackets please, especially as the placement of
> brackets affects the result of this particular calculation.
Or simply eliminate one of the divisions using
"1000000ull / APIC_BUS_CYCLE_NS".
>> --- a/xen/arch/x86/traps.c
>> +++ b/xen/arch/x86/traps.c
>> @@ -685,8 +685,12 @@ int cpuid_hypervisor_leaves( uint32_t idx, uint32_t sub_idx,
>> uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx)
>> {
>> struct domain *d = current->domain;
>> - /* Optionally shift out of the way of Viridian architectural leaves. */
>> - uint32_t base = is_viridian_domain(d) ? 0x40000100 : 0x40000000;
>> + /*
>> + * Optionally shift out of the way of Viridian or VMware
>> + * architectural leaves.
>> + */
>> + uint32_t base = is_viridian_domain(d) | is_vmware_domain(d) ?
>> + 0x40000100 : 0x40000000;
>
> Again, brackets please for binary operators. (We have had one recent
> slipup because of the precedence of | and ?:)
I don't think parentheses are strictly required here - we're not very
consistent with when to disambiguate operator precedence by using
them when not strictly needed anyway, and I think people touching
hypervisor code can be expected to know the language specification
well enough.
Jan
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-11 18:36 ` [PATCH v4 01/16] xen: Add support for VMware cpuid leaves Don Slutz
2014-09-11 19:49 ` Andrew Cooper
@ 2014-09-12 12:37 ` Boris Ostrovsky
2014-09-12 17:50 ` Don Slutz
1 sibling, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-12 12:37 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/11/2014 02:36 PM, Don Slutz wrote:
> diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
> index 10fc2ca..f353f42 100644
> --- a/xen/arch/x86/traps.c
> +++ b/xen/arch/x86/traps.c
> @@ -685,8 +685,12 @@ int cpuid_hypervisor_leaves( uint32_t idx, uint32_t sub_idx,
> uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx)
> {
> struct domain *d = current->domain;
> - /* Optionally shift out of the way of Viridian architectural leaves. */
> - uint32_t base = is_viridian_domain(d) ? 0x40000100 : 0x40000000;
> + /*
> + * Optionally shift out of the way of Viridian or VMware
> + * architectural leaves.
> + */
> + uint32_t base = is_viridian_domain(d) | is_vmware_domain(d) ?
Given how is_viridian and is_vmware are defined I think '||' is more
appropriate.
-boris
> + 0x40000100 : 0x40000000;
> uint32_t limit, dummy;
>
> idx -= base;
> diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
> index 1123857..546210a 100644
> --- a/xen/include/asm-x86/hvm/hvm.h
> +++ b/xen/include/asm-x86/hvm/hvm.h
> @@ -347,6 +347,9 @@ static inline unsigned long hvm_get_shadow_gs_base(struct vcpu *v)
> #define is_viridian_domain(_d) \
> (is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VIRIDIAN]))
>
> +#define is_vmware_domain(_d) \
> + (is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW]))
> +
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-11 18:36 ` [PATCH v4 04/16] xen: Add is_vmware_port_enabled Don Slutz
2014-09-11 21:22 ` Andrew Cooper
@ 2014-09-12 13:08 ` Boris Ostrovsky
2014-09-16 12:08 ` Slutz, Donald Christopher
1 sibling, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-12 13:08 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/11/2014 02:36 PM, Don Slutz wrote:
> int __get_instruction_length_from_list(struct vcpu *v,
> - const enum instruction_index *list, unsigned int list_count)
> + const enum instruction_index *list,
> + unsigned int list_count,
> + bool_t err_rpt)
> {
> struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
> unsigned int i, j, inst_len = 0;
> @@ -211,10 +222,13 @@ int __get_instruction_length_from_list(struct vcpu *v,
> mismatch: ;
> }
>
> - gdprintk(XENLOG_WARNING,
> - "%s: Mismatch between expected and actual instruction bytes: "
> - "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
> - hvm_inject_hw_exception(TRAP_gp_fault, 0);
> + if ( err_rpt )
> + {
> + gdprintk(XENLOG_WARNING,
> + "%s: Mismatch between expected and actual instruction bytes: "
> + "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
> + hvm_inject_hw_exception(TRAP_gp_fault, 0);
> + }
> return 0;
>
> done:
> diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
> index b5188e6..9e14d2a 100644
> --- a/xen/arch/x86/hvm/svm/svm.c
> +++ b/xen/arch/x86/hvm/svm/svm.c
> @@ -59,6 +59,7 @@
> #include <public/sched.h>
> #include <asm/hvm/vpt.h>
> #include <asm/hvm/trace.h>
> +#include <asm/hvm/vmport.h>
> #include <asm/hap.h>
> #include <asm/apic.h>
> #include <asm/debugger.h>
> @@ -2065,6 +2066,38 @@ svm_vmexit_do_vmsave(struct vmcb_struct *vmcb,
> return;
> }
>
> +static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
> + struct vcpu *v)
> +{
> + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
> + unsigned long inst_len;
> + unsigned long inst_addr = svm_rip2pointer(v);
> + int rc;
> + static const enum instruction_index list[] = {
> + INSTR_INL_DX, INSTR_INB_DX, INSTR_OUTL_DX, INSTR_OUTB_DX
> + };
> +
> + inst_len = __get_instruction_length_from_list(
> + v, list, ARRAY_SIZE(list), 0);
I should have asked earlier but I don't think I understand why the last
argument here is 0 (and therefore why you have this last argument at all).
Because whether or not you are warning in
__get_instruction_length_from_list() it will still return 0. And that,
in turn, will cause vmport_gp_check() to return an error. And then you
will print another warning in VMPORT_LOG. So there is a warning anyway.
Second, since this handler appears to be handling #GP only for VMware
guest we should make sure that it is not executed for any other guest.
You do now condition intercept got #GP for such guests only but I still
think having a check here is worth doing. Maybe a BUG() or ASSERT()?
The same comments are applicable to VMX code, I suspect.
> +
> + rc = vmport_gp_check(regs, v, inst_len, inst_addr,
> + vmcb->exitinfo1, vmcb->exitinfo2);
> + if ( !rc )
> + __update_guest_eip(regs, inst_len);
> + else
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_GP_UNKNOWN,
> + "gp: rc=%d ei1=0x%lx ei2=0x%lx ip=%"PRIx64
> + " (0x%lx,%ld) ax=%"PRIx64" bx=%"PRIx64" cx=%"PRIx64
> + " dx=%"PRIx64" si=%"PRIx64" di=%"PRIx64, rc,
> + (unsigned long)vmcb->exitinfo1,
> + (unsigned long)vmcb->exitinfo2, regs->rip, inst_addr,
> + inst_len, regs->rax, regs->rbx, regs->rcx, regs->rdx,
> + regs->rsi, regs->rdi);
> + hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
> + }
> +}
> +
...
> +
> +int vmport_gp_check(struct cpu_user_regs *regs, struct vcpu *v,
> + unsigned long inst_len, unsigned long inst_addr,
> + unsigned long ei1, unsigned long ei2)
> +{
> + if ( !v->domain->arch.hvm_domain.is_vmware_port_enabled )
> + return 10;
> +
> + if ( inst_len && inst_len <= 2 && get_low_bits(regs->rdx) == BDOOR_PORT &&
> + ei1 == 0 && ei2 == 0 && regs->error_code == 0 &&
> + (uint32_t)regs->rax == BDOOR_MAGIC )
> + {
> + int i = 0;
> + uint32_t val;
> + uint32_t byte_cnt = 4;
> + unsigned char bytes[2];
> + unsigned int fetch_len;
> + int frc;
> + int rc;
> +
> + /*
> + * Fetch up to the next page break; we'll fetch from the
> + * next page later if we have to.
> + */
> + fetch_len = min_t(unsigned int, inst_len,
> + PAGE_SIZE - (inst_addr & ~PAGE_MASK));
> + frc = hvm_fetch_from_guest_virt_nofault(bytes, inst_addr, fetch_len,
> + PFEC_page_present);
> + if ( frc != HVMCOPY_okay )
> + {
> + gdprintk(XENLOG_WARNING,
> + "Bad instruction fetch at %#lx (frc=%d il=%lu fl=%u)\n",
> + (unsigned long) inst_addr, frc, inst_len, fetch_len);
> + return 11;
> + }
> + if ( bytes[0] == 0x66 ) /* operand size prefix */
> + {
> + byte_cnt = 2;
> + i = 1;
> + if ( fetch_len != inst_len )
> + {
> + frc = hvm_fetch_from_guest_virt_nofault(&bytes[1],
> + inst_addr + 1, 1,
> + PFEC_page_present);
> + if ( frc != HVMCOPY_okay )
> + {
> + gdprintk(XENLOG_WARNING,
> + "Bad instruction fetch at %#lx + 1 (frc=%d)\n",
> + (unsigned long) inst_addr, frc);
> + return 12;
> + }
> + }
> + }
> + if ( bytes[i] == 0xed ) /* in (%dx),%eax or in (%dx),%ax */
> + {
> + rc = vmport_ioport(IOREQ_READ, BDOOR_PORT, byte_cnt, &val);
> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
> + "gp: VMwareIn rc=%d ip=%"PRIx64" byte_cnt=%d ax=%"
> + PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
> + " si=%"PRIx64" di=%"PRIx64, rc,
> + inst_addr, byte_cnt, regs->rax, regs->rbx,
> + regs->rcx, regs->rdx, regs->rsi, regs->rdi);
> + return rc;
> + }
> + else if ( bytes[i] == 0xec ) /* in (%dx),%al */
> + {
> + rc = vmport_ioport(IOREQ_READ, BDOOR_PORT, 1, &val);
> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
> + "gp: VMwareIn rc=%d ip=%"PRIx64" byte_cnt=1 ax=%"
> + PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
> + " si=%"PRIx64" di=%"PRIx64, rc,
> + inst_addr, regs->rax, regs->rbx, regs->rcx,
> + regs->rdx, regs->rsi, regs->rdi);
> + return rc;
> + }
> + else if ( bytes[i] == 0xef ) /* out %eax,(%dx) or out %ax,(%dx) */
> + {
> + rc = vmport_ioport(IOREQ_WRITE, BDOOR_PORT, byte_cnt, &val);
> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
> + "gp: VMwareOut rc=%d ip=%"PRIx64" byte_cnt=%d ax=%"
> + PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
> + " si=%"PRIx64" di=%"PRIx64, rc,
> + inst_addr, byte_cnt, regs->rax, regs->rbx,
> + regs->rcx, regs->rdx, regs->rsi, regs->rdi);
> + return rc;
> + }
> + else if ( bytes[i] == 0xee ) /* out %al,(%dx) */
> + {
> + rc = vmport_ioport(IOREQ_WRITE, BDOOR_PORT, 1, &val);
> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
> + "gp: VMwareOut rc=%d ip=%"PRIx64" byte_cnt=1 ax=%"
> + PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
> + " si=%"PRIx64" di=%"PRIx64, rc,
> + inst_addr, regs->rax, regs->rbx, regs->rcx,
> + regs->rdx, regs->rsi, regs->rdi);
> + return rc;
> + }
> + else
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_GP_FAIL_RD_INST,
> + "gp: VMware? lip=%"PRIx64"[%d]=>0x%x(%ld) ax=%"
> + PRIx64" bx=%"PRIx64" cx=%"PRIx64" dx=%"PRIx64
> + " si=%"PRIx64" di=%"PRIx64,
> + inst_addr, i, bytes[i], inst_len, regs->rax,
> + regs->rbx, regs->rcx, regs->rdx, regs->rsi,
> + regs->rdi);
> + return 13;
> + }
> + }
> + return 14;
The return values should be defined as macros --- otherwise they look
like some magic integers.
-boris
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 06/16] xen: Convert vmware_port to xentrace usage
2014-09-11 18:36 ` [PATCH v4 06/16] xen: Convert vmware_port to xentrace usage Don Slutz
@ 2014-09-12 13:10 ` Boris Ostrovsky
2014-09-12 23:57 ` Don Slutz
0 siblings, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-12 13:10 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/11/2014 02:36 PM, Don Slutz wrote:
> @@ -2080,6 +2080,14 @@ static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
> inst_len = __get_instruction_length_from_list(
> v, list, ARRAY_SIZE(list), 0);
>
> + if ( hvm_long_mode_enabled(v) )
> + HVMTRACE_LONG2_C4D(TRAP_GP, inst_len, regs->rdx,
> + TRC_PAR_LONG(vmcb->exitinfo1),
> + TRC_PAR_LONG(vmcb->exitinfo2));
> + else
> + HVMTRACE_C4D(TRAP_GP, inst_len, vmcb->exitinfo1, vmcb->exitinfo1,
> + vmcb->exitinfo2);
exitinfo1 is used twice. Did you mean regs->rdx for one of them?
-boris
> +
> rc = vmport_gp_check(regs, v, inst_len, inst_addr,
> vmcb->exitinfo1, vmcb->exitinfo2);
> if ( !rc )
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 07/16] tools: Convert vmware_port to xentrace usage
2014-09-11 18:36 ` [PATCH v4 07/16] tools: " Don Slutz
@ 2014-09-12 13:15 ` Boris Ostrovsky
2014-09-13 0:01 ` Don Slutz
0 siblings, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-12 13:15 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/11/2014 02:36 PM, Don Slutz wrote:
> Also added missing TRAP_DEBUG & VLAPIC.
>
> Signed-off-by: Don Slutz <dslutz@verizon.com>
> ---
> tools/xentrace/formats | 13 +++++++++++++
> 1 file changed, 13 insertions(+)
>
> diff --git a/tools/xentrace/formats b/tools/xentrace/formats
> index da658bf..b403c54 100644
> --- a/tools/xentrace/formats
> +++ b/tools/xentrace/formats
> @@ -79,6 +79,19 @@
> 0x00082020 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) INTR_WINDOW [ value = 0x%(1)08x ]
> 0x00082021 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) NPF [ gpa = 0x%(2)08x%(1)08x mfn = 0x%(4)08x%(3)08x qual = 0x%(5)04x p2mt = 0x%(6)04x ]
> 0x00082023 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP [ vector = 0x%(1)02x ]
> +0x00082024 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_DEBUG [ exit_qualification = 0x%(1)08x ]
> +0x00082025 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VLAPIC
> +0x00082026 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP [ inst_len = %(1)d edx = 0x%(2)08x exitinfo1 = 0x%(3)08x exitinfo2 = 0x%(4)08x ]
> +0x00082126 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP [ inst_len = %(1)d edx = 0x%(2)08x exitinfo1 = 0x%(4)08x%(3)08x exitinfo2 = 0x%(6)08x%(5)08x ]
> +0x00082027 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP_UNKNOWN [ rc = %(1)d eax = 0x%(2)08x ebx = 0x%(3)08x ecx = 0x%(4)08x inst_addr = 0x%(5)08x ]
> +0x00082127 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP_UNKNOWN [ rc = %(1)d eax = 0x%(2)08x ebx = 0x%(3)08x ecx = 0x%(4)08x inst_addr = 0x%(6)08x%(5)08x ]
> +0x00082028 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_READ_BEFORE [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
> +0x00082228 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_WRITE_BEFORE [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
> +0x00082029 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_READ_AFTER [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
> +0x00082229 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_WRITE_AFTER [ eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
> +0x0008202a CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_BAD [ dir = %(1)d bytes = 0x%(2)d eax = 0x%(3)08x eip = 0x%(4)08x ]
'bytes = 0x%(2)d' or 'bytes = %(2)d' ?
-boris
> +0x0008212a CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_BAD [ dir = %(1)d bytes = %(2)d eax = 0x%(3)08x rip = 0x%(5)08x%(4)08x ]
> +0x0008202b CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_UNKNOWN [ bytes << 8 + dir = 0x%(1)03x cmd = 0x%(2)x cmd = %(2)d ebx = 0x%(3)08x ecx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
>
> 0x0010f001 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) page_grant_map [ domid = %(1)d ]
> 0x0010f002 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) page_grant_unmap [ domid = %(1)d ]
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc
2014-09-11 18:36 ` [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc Don Slutz
@ 2014-09-12 13:37 ` Boris Ostrovsky
2014-09-12 14:27 ` Jan Beulich
2014-09-16 13:17 ` Slutz, Donald Christopher
0 siblings, 2 replies; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-12 13:37 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/11/2014 02:36 PM, Don Slutz wrote:
> @@ -6142,6 +6148,43 @@ long do_hvm_op(unsigned long op, XEN_GUEST_HANDLE_PARAM(void) arg)
> break;
> }
>
> + case HVMOP_get_vmport_guest_info:
> + case HVMOP_set_vmport_guest_info:
> + {
> + struct xen_hvm_vmport_guest_info a;
> + struct domain *d;
> +
> + if ( copy_from_guest(&a, arg, 1) )
> + return -EFAULT;
> +
> + rc = vmport_rpc_hvmop_precheck(op, &a);
> + if ( rc )
> + return rc;
> +
> + d = rcu_lock_domain_by_any_id(a.domid);
> + if ( d == NULL )
> + return rc;
> +
> + rc = -EINVAL;
> + if ( !is_hvm_domain(d) )
> + goto param_fail9;
Do we need this check? If you are concerned about PVH then checking
explicitly for that may be better (and probably do it first thing in the
case clause)
> +
> + rc = xsm_hvm_param(XSM_TARGET, d, op);
> + if ( rc )
> + goto param_fail9;
> +
> + rc = vmport_rpc_hvmop_do(d, op, &a);
> + if ( rc )
> + goto param_fail9;
> +
> + if ( op == HVMOP_get_vmport_guest_info )
> + rc = copy_to_guest(arg, &a, 1) ? -EFAULT : 0;
> +
> + param_fail9:
> + rcu_unlock_domain(d);
> + break;
> + }
> +
> default:
> {
> gdprintk(XENLOG_DEBUG, "Bad HVM op %ld.\n", op);
> diff --git a/xen/arch/x86/hvm/vmware/Makefile b/xen/arch/x86/hvm/vmware/Makefile
> index cd8815b..4a14124 100644
> --- a/xen/arch/x86/hvm/vmware/Makefile
> +++ b/xen/arch/x86/hvm/vmware/Makefile
> @@ -1,2 +1,3 @@
> obj-y += cpuid.o
> obj-y += vmport.o
> +obj-y += vmport_rpc.o
> diff --git a/xen/arch/x86/hvm/vmware/vmport.c b/xen/arch/x86/hvm/vmware/vmport.c
> index 26aeb37..9e308a0 100644
> --- a/xen/arch/x86/hvm/vmware/vmport.c
> +++ b/xen/arch/x86/hvm/vmware/vmport.c
> @@ -139,6 +139,13 @@ int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val)
> /* maxTimeLag */
> regs->rcx = 0;
> break;
> + case BDOOR_CMD_MESSAGE:
> + if ( !is_pv_vcpu(current) )
> + {
> + /* Only supported for non pv domains */
PV vs. HVM vs. PVH. So probably 'if(is_hvm_vcpu)'?
> + vmport_rpc(¤t->domain->arch.hvm_domain, regs);
> + }
> + break;
> case BDOOR_CMD_GETGUIOPTIONS:
> regs->rax = VMWARE_GUI_AUTO_GRAB | VMWARE_GUI_AUTO_UNGRAB |
> VMWARE_GUI_AUTO_RAISE_DISABLED | VMWARE_GUI_SYNC_TIME |
> diff --git a/xen/arch/x86/hvm/vmware/vmport_rpc.c b/xen/arch/x86/hvm/vmware/vmport_rpc.c
> new file mode 100644
> index 0000000..695e58f
> --- /dev/null
> +++ b/xen/arch/x86/hvm/vmware/vmport_rpc.c
> @@ -0,0 +1,1281 @@
> +/*
> + * HVM VMPORT RPC emulation
> + *
> + * Copyright (C) 2012 Verizon Corporation
> + *
> + * This file is free software; you can redistribute it and/or modify it
> + * under the terms of the GNU General Public License Version 2 (GPLv2)
> + * as published by the Free Software Foundation.
> + *
> + * This file is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + * General Public License for more details. <http://www.gnu.org/licenses/>.
> + */
> +
> +/*
> + * VMware Tools running in a DOMU will do "info-get" and "info-set"
> + * guestinfo commands to get and set keys and values. Inside the VM,
> + * vmtools at its lower level will feed the command string 4 bytes
> + * at a time into the VMWARE magic port using the IN
> + * instruction. Each 4 byte mini-rpc will get handled
> + * vmport_io()-->vmport_rpc()-->vmport_process_packet()-->
> + * vmport_process_send_payload()-->vmport_send() and the command
> + * string will get accumulated into a channels send_buffer. When
> + * the full length of the string has been accumulated, then this
> + * code copies the send_buffer into a free
> + * vmport_state->channel-->receive_bucket.buffer
> + * VMware tools then does RECVSIZE and RECVPAYLOAD messages, the
> + * latter then reads 4 bytes at a time using the IN instruction (for
> + * the info-get case). Then a final RECVSTATUS message is sent to
> + * finish up
> + */
> +
> +#include <xen/config.h>
> +#include <xen/lib.h>
> +#include <xen/cper.h>
> +#include <asm/hvm/hvm.h>
> +#include <asm/hvm/support.h>
> +#include <asm/hvm/vmport.h>
> +#include <asm/hvm/trace.h>
> +
> +#include "public/arch-x86/hvm/vmporttype.h"
> +
> +#include "backdoor_def.h"
> +#include "guest_msg_def.h"
> +
> +
> +#define VMWARE_PROTO_TO_GUEST 0x4f4c4354
> +#define VMWARE_PROTO_FROM_GUEST 0x49435052
> +
> +#define GUESTINFO_NOTFOUND 500
> +#define GUESTINFO_VALTOOLONG 1
> +#define GUESTINFO_KEYTOOLONG 2
> +#define GUESTINFO_TOOMANYKEYS 3
> +
> +
> +static inline uint16_t get_low_bits(uint32_t bits)
> +{
> + return bits & 0xffff;
> +}
> +
> +static inline uint16_t get_high_bits(uint32_t bits)
> +{
> + return bits >> 16;
> +}
> +
> +static inline uint32_t set_high_bits(uint32_t b, uint32_t val)
> +{
> + return (val << 16) | get_low_bits(b);
> +}
> +
> +static inline void set_status(struct cpu_user_regs *ur, uint16_t val)
> +{
> + /* VMware defines this to be only 32 bits */
> + ur->rcx = (val << 16) | (ur->rcx & 0xffff);
Are you shifting val out of its size? Or will compiler automatically
promote this to uint32_t? (It does for me but I don't know if this is
something we can rely on).
-boris
> +}
> +
> +#ifndef NDEBUG
> +static void vmport_safe_print(char *prefix, int len, const char *msg)
> +{
> + unsigned char c;
> + unsigned int end = len;
> + unsigned int i, k;
> + char out[4 * (VMPORT_MAX_SEND_BUF + 1) * 3 + 6];
> +
> + if ( end > (sizeof(out) / 3 - 6) )
> + end = sizeof(out) / 3 - 6;
> + out[0] = '<';
> + k = 1;
> + for ( i = 0; i < end; i++ )
> + {
> + c = msg[i];
> + if ( (c == '^') || (c == '\\') || (c == '>') )
> + {
> + out[k++] = '\\';
> + out[k++] = c;
> + }
> + else if ( (c >= ' ') && (c <= '~') )
> + out[k++] = c;
> + else if ( c < ' ' )
> + {
> + out[k++] = '^';
> + out[k++] = c ^ 0x40;
> + }
> + else
> + {
> + snprintf(&out[k], sizeof(out) - k, "\\%02x", c);
> + k += 3;
> + }
> + }
> + out[k++] = '>';
> + if ( len > end )
> + {
> + out[k++] = '.';
> + out[k++] = '.';
> + out[k++] = '.';
> + }
> + out[k++] = 0;
> + gdprintk(XENLOG_DEBUG, "%s%d(%d,%d,%zu)%s\n", prefix, end, len, k,
> + sizeof(out), out);
> +}
> +#endif
> +
> +/*
> + * Copy message into a jumbo bucket buffer which vmtools will use to
> + * read from 4 bytes at a time until done with it
> + */
> +static void vmport_send_jumbo(struct hvm_domain *hd, vmport_channel_t *c,
> + const char *msg)
> +{
> + unsigned int cur_recv_len = strlen(msg) + 1;
> + vmport_jumbo_bucket_t *b = &(c->jumbo_recv_bkt);
> +
> + b->ctl.recv_len = cur_recv_len;
> + b->ctl.recv_idx = 0;
> +
> + memset(b->recv_buf, 0, sizeof(b->recv_buf));
> +
> + if ( cur_recv_len >= (sizeof(b->recv_buf) - 1) )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
> + "VMware jumbo recv_len=%d >= %ld",
> + cur_recv_len, sizeof(b->recv_buf) - 1);
> + cur_recv_len = sizeof(b->recv_buf) - 1;
> + }
> +
> + memcpy(b->recv_buf, msg, cur_recv_len);
> +
> + c->ctl.jumbo = 1;
> +}
> +
> +/*
> + * Copy message into a free receive bucket buffer which vmtools will use to
> + * read from 4 bytes at a time until done with it
> + */
> +static void vmport_send_normal(struct hvm_domain *hd, vmport_channel_t *c,
> + const char *msg)
> +{
> + unsigned int cur_recv_len = strlen(msg) + 1;
> + unsigned int my_bkt = c->ctl.recv_write;
> + unsigned int next_bkt = my_bkt + 1;
> + vmport_bucket_t *b;
> +
> + if ( next_bkt >= VMPORT_MAX_BKTS )
> + next_bkt = 0;
> +
> + if ( next_bkt == c->ctl.recv_read )
> + {
> +#ifndef NDEBUG
> + if ( opt_vmport_debug & VMPORT_LOG_SKIP_SEND )
> + {
> + char prefix[30];
> +
> + snprintf(prefix, sizeof(prefix),
> + "VMware _send skipped %d (%d, %d) ",
> + c->ctl.chan_id, my_bkt, c->ctl.recv_read);
> + prefix[sizeof(prefix) - 1] = 0;
> + vmport_safe_print(prefix, cur_recv_len, msg);
> + }
> +#endif
> + return;
> + }
> +
> + c->ctl.recv_write = next_bkt;
> + b = &c->recv_bkt[my_bkt];
> +#ifndef NDEBUG
> + if ( opt_vmport_debug & VMPORT_LOG_SEND )
> + {
> + char prefix[30];
> +
> + snprintf(prefix, sizeof(prefix), "VMware _send %d (%d) ",
> + c->ctl.chan_id, my_bkt);
> + prefix[sizeof(prefix) - 1] = 0;
> + vmport_safe_print(prefix, cur_recv_len, msg);
> + }
> +#endif
> +
> + b->ctl.recv_len = cur_recv_len;
> + b->ctl.recv_idx = 0;
> + memset(b->recv_buf, 0, sizeof(b->recv_buf));
> + if ( cur_recv_len >= (sizeof(b->recv_buf) - 1) )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware recv_len=%d >= %zd",
> + cur_recv_len, sizeof(b->recv_buf) - 1);
> + cur_recv_len = sizeof(b->recv_buf) - 1;
> + }
> + memcpy(b->recv_buf, msg, cur_recv_len);
> +}
> +
> +static void vmport_send(struct hvm_domain *hd, vmport_channel_t *c,
> + const char *msg)
> +{
> + unsigned int cur_recv_len = strlen(msg) + 1;
> +
> + if ( cur_recv_len > VMPORT_MAX_VAL_LEN )
> + vmport_send_jumbo(hd, c, msg);
> + else
> + vmport_send_normal(hd, c, msg);
> +}
> +
> +void vmport_ctrl_send(struct hvm_domain *hd, char *msg)
> +{
> + struct vmport_state *vs = hd->vmport_data;
> + unsigned int i;
> +
> + hd->vmport_data->ping_time = get_sec();
> + spin_lock(&hd->vmport_lock);
> + for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
> + {
> + if ( vs->chans[i].ctl.proto_num == VMWARE_PROTO_TO_GUEST )
> + vmport_send(hd, &vs->chans[i], msg);
> + }
> + spin_unlock(&hd->vmport_lock);
> +}
> +
> +static void vmport_flush(struct hvm_domain *hd)
> +{
> + spin_lock(&hd->vmport_lock);
> + memset(&hd->vmport_data->chans, 0, sizeof(hd->vmport_data->chans));
> + spin_unlock(&hd->vmport_lock);
> +}
> +
> +static void vmport_sweep(struct hvm_domain *hd, unsigned long now_time)
> +{
> + struct vmport_state *vs = hd->vmport_data;
> + unsigned int i;
> +
> + for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
> + {
> + if ( vs->chans[i].ctl.proto_num )
> + {
> + vmport_channel_t *c = &vs->chans[i];
> + long delta = now_time - c->ctl.active_time;
> +
> + if ( delta >= 80 )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_SWEEP, "VMware flush %d. delta=%ld",
> + c->ctl.chan_id, delta);
> + /* Return channel to free pool */
> + c->ctl.proto_num = 0;
> + }
> + }
> + }
> +}
> +
> +static vmport_channel_t *vmport_new_chan(struct vmport_state *vs,
> + unsigned long now_time)
> +{
> + unsigned int i;
> +
> + for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
> + {
> + if ( !vs->chans[i].ctl.proto_num )
> + {
> + vmport_channel_t *c = &vs->chans[i];
> +
> + c->ctl.chan_id = i;
> + c->ctl.cookie = vs->open_cookie++;
> + c->ctl.active_time = now_time;
> + c->ctl.send_len = 0;
> + c->ctl.send_idx = 0;
> + c->ctl.recv_read = 0;
> + c->ctl.recv_write = 0;
> + return c;
> + }
> + }
> + return NULL;
> +}
> +
> +static void vmport_process_send_size(struct hvm_domain *hd, vmport_channel_t *c,
> + struct cpu_user_regs *ur)
> +{
> + /* vmware tools often send a 0 byte request size. */
> + c->ctl.send_len = ur->rbx;
> + c->ctl.send_idx = 0;
> +
> + set_status(ur, MESSAGE_STATUS_SUCCESS);
> +}
> +
> +/* ret_buffer is in/out param */
> +static int vmport_get_guestinfo(struct hvm_domain *hd, struct vmport_state *vs,
> + char *a_info_key, unsigned int a_key_len,
> + char *ret_buffer, unsigned int ret_buffer_len)
> +{
> + unsigned int i;
> +
> + for ( i = 0; i < vs->used_guestinfo; i++ )
> + {
> + if ( vs->guestinfo[i] &&
> + (vs->guestinfo[i]->key_len == a_key_len) &&
> + (memcmp(a_info_key, vs->guestinfo[i]->key_data,
> + vs->guestinfo[i]->key_len) == 0) )
> + {
> + snprintf(ret_buffer, ret_buffer_len - 1, "1 %.*s",
> + (int)vs->guestinfo[i]->val_len,
> + vs->guestinfo[i]->val_data);
> + return i;
> + }
> + }
> +
> + for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
> + {
> + if ( vs->guestinfo_jumbo[i] &&
> + (vs->guestinfo_jumbo[i]->key_len == a_key_len) &&
> + (memcmp(a_info_key, vs->guestinfo_jumbo[i]->key_data,
> + vs->guestinfo_jumbo[i]->key_len) == 0) )
> + {
> + snprintf(ret_buffer, ret_buffer_len - 1, "1 %.*s",
> + (int)vs->guestinfo_jumbo[i]->val_len,
> + vs->guestinfo_jumbo[i]->val_data);
> + return i;
> + }
> + }
> + return GUESTINFO_NOTFOUND;
> +}
> +
> +static void hvm_del_guestinfo_jumbo(struct vmport_state *vs, char *key,
> + uint8_t len)
> +{
> + int i;
> +
> + for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
> + {
> + if ( !vs->guestinfo_jumbo[i] )
> + {
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING,
> + "i=%d not allocated used_guestinfo_jumbo=%d\n",
> + i, vs->used_guestinfo_jumbo);
> +#endif
> + }
> + else if ( (vs->guestinfo_jumbo[i]->key_len == len) &&
> + (memcmp(key, vs->guestinfo_jumbo[i]->key_data, len) == 0) )
> + {
> + vs->guestinfo_jumbo[i]->key_len = 0;
> + vs->guestinfo_jumbo[i]->val_len = 0;
> + break;
> + }
> + }
> +}
> +
> +static void hvm_del_guestinfo(struct vmport_state *vs, char *key, uint8_t len)
> +{
> + int i;
> +
> + for ( i = 0; i < vs->used_guestinfo; i++ )
> + {
> + if ( !vs->guestinfo[i] )
> + {
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING,
> + "i=%d not allocated, but used_guestinfo=%d\n",
> + i, vs->used_guestinfo);
> +#endif
> + }
> + else if ( (vs->guestinfo[i]->key_len == len) &&
> + (memcmp(key, vs->guestinfo[i]->key_data, len) == 0) )
> + {
> + vs->guestinfo[i]->key_len = 0;
> + vs->guestinfo[i]->val_len = 0;
> + break;
> + }
> + }
> +}
> +
> +static int vmport_set_guestinfo(struct vmport_state *vs, int a_key_len,
> + unsigned int a_val_len, char *a_info_key, char *val)
> +{
> + unsigned int i;
> + int free_i = -1, rc = 0;
> +
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING, "vmport_set_guestinfo a_val_len=%d\n", a_val_len);
> +#endif
> +
> + if ( a_key_len <= VMPORT_MAX_KEY_LEN )
> + {
> + if ( a_val_len <= VMPORT_MAX_VAL_LEN )
> + {
> + for ( i = 0; i < vs->used_guestinfo; i++ )
> + {
> + if ( !vs->guestinfo[i] )
> + {
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING,
> + "i=%d not allocated, but used_guestinfo=%d\n",
> + i, vs->used_guestinfo);
> +#endif
> + }
> + else if ( (vs->guestinfo[i]->key_len == a_key_len) &&
> + (memcmp(a_info_key, vs->guestinfo[i]->key_data,
> + vs->guestinfo[i]->key_len) == 0) )
> + {
> + vs->guestinfo[i]->val_len = a_val_len;
> + memcpy(vs->guestinfo[i]->val_data, val, a_val_len);
> + break;
> + }
> + else if ( (vs->guestinfo[i]->key_len == 0) &&
> + (free_i == -1) )
> + free_i = i;
> + }
> + if ( i >= vs->used_guestinfo )
> + {
> + if ( free_i == -1 )
> + rc = GUESTINFO_TOOMANYKEYS;
> + else
> + {
> + vs->guestinfo[free_i]->key_len = a_key_len;
> + memcpy(vs->guestinfo[free_i]->key_data,
> + a_info_key, a_key_len);
> + vs->guestinfo[free_i]->val_len = a_val_len;
> + memcpy(vs->guestinfo[free_i]->val_data,
> + val, a_val_len);
> + }
> + }
> + }
> + else
> + rc = GUESTINFO_VALTOOLONG;
> + }
> + else
> + rc = GUESTINFO_KEYTOOLONG;
> + if ( !rc )
> + hvm_del_guestinfo_jumbo(vs, a_info_key, a_key_len);
> + return rc;
> +}
> +
> +static int vmport_set_guestinfo_jumbo(struct vmport_state *vs, int a_key_len,
> + int a_val_len, char *a_info_key, char *val)
> +{
> + unsigned int i;
> + int free_i = -1, rc = 0;
> +
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING, "vmport_set_guestinfo_jumbo a_val_len=%d\n",
> + a_val_len);
> +#endif
> +
> + if ( a_key_len <= VMPORT_MAX_KEY_LEN )
> + {
> + if ( a_val_len <= VMPORT_MAX_VAL_JUMBO_LEN )
> + {
> + for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
> + {
> + if ( !vs->guestinfo_jumbo[i] )
> + {
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING,
> + "i=%d not allocated; used_guestinfo_jumbo=%d\n",
> + i, vs->used_guestinfo_jumbo);
> +#endif
> + }
> + else if ( (vs->guestinfo_jumbo[i]->key_len == a_key_len) &&
> + (memcmp(a_info_key,
> + vs->guestinfo_jumbo[i]->key_data,
> + vs->guestinfo_jumbo[i]->key_len) == 0) )
> + {
> +
> + vs->guestinfo_jumbo[i]->val_len = a_val_len;
> + memcpy(vs->guestinfo_jumbo[i]->val_data, val, a_val_len);
> + break;
> + }
> + else if ( (vs->guestinfo_jumbo[i]->key_len == 0) &&
> + (free_i == -1) )
> + free_i = i;
> + }
> + if ( i >= vs->used_guestinfo_jumbo )
> + {
> + if ( free_i == -1 )
> + rc = GUESTINFO_TOOMANYKEYS;
> + else
> + {
> + vs->guestinfo_jumbo[free_i]->key_len = a_key_len;
> + memcpy(vs->guestinfo_jumbo[free_i]->key_data,
> + a_info_key, a_key_len);
> + vs->guestinfo_jumbo[free_i]->val_len = a_val_len;
> + memcpy(vs->guestinfo_jumbo[free_i]->val_data,
> + val, a_val_len);
> + }
> + }
> + }
> + else
> + rc = GUESTINFO_VALTOOLONG;
> + }
> + else
> + rc = GUESTINFO_KEYTOOLONG;
> + if ( !rc )
> + hvm_del_guestinfo(vs, a_info_key, a_key_len);
> + return rc;
> +}
> +
> +static void vmport_process_send_payload(struct hvm_domain *hd,
> + vmport_channel_t *c,
> + struct cpu_user_regs *ur,
> + unsigned long now_time)
> +{
> + /* Accumulate 4 bytes of paload into send_buf using offset */
> + if ( c->ctl.send_idx < VMPORT_MAX_SEND_BUF )
> + c->send_buf[c->ctl.send_idx] = ur->rbx;
> +
> + c->ctl.send_idx++;
> + set_status(ur, MESSAGE_STATUS_SUCCESS);
> +
> + if ( c->ctl.send_idx * 4 >= c->ctl.send_len )
> + {
> +
> + /* We are done accumulating so handle the command */
> +
> + if ( c->ctl.send_idx < VMPORT_MAX_SEND_BUF )
> + ((char *)c->send_buf)[c->ctl.send_len] = 0;
> +#ifndef NDEBUG
> + if ( opt_vmport_debug & VMPORT_LOG_RECV )
> + {
> + char prefix[30];
> +
> + snprintf(prefix, sizeof(prefix),
> + "VMware RECV %d (%d) ", c->ctl.chan_id, c->ctl.recv_read);
> + prefix[sizeof(prefix) - 1] = 0;
> + vmport_safe_print(prefix, c->ctl.send_len, (char *)c->send_buf);
> + }
> +#endif
> + if ( c->ctl.proto_num == VMWARE_PROTO_FROM_GUEST )
> + {
> + /*
> + * Eaxmples of messages:
> + *
> + * log toolbox: Version: build-341836
> + * SetGuestInfo 4 build-341836
> + * info-get guestinfo.ip
> + * info-set guestinfo.ip joe
> + *
> + */
> +
> + char *build = NULL;
> + char *info_key = NULL;
> + char *ret_msg = "1 ";
> + char ret_buffer[2 + VMPORT_MAX_VAL_JUMBO_LEN + 2];
> +
> + if ( strncmp((char *)c->send_buf, "log toolbox: Version: build-",
> + strlen("log toolbox: Version: build-")) == 0 )
> +
> + build = (char *)c->send_buf +
> + strlen("log toolbox: Version: build-");
> +
> + else if ( strncmp((char *)c->send_buf, "SetGuestInfo 4 build-",
> + strlen("SetGuestInfo 4 build-")) == 0 )
> +
> + build = (char *)c->send_buf + strlen("SetGuestInfo 4 build-");
> +
> + else if ( strncmp((char *)c->send_buf, "info-get guestinfo.",
> + strlen("info-get guestinfo.")) == 0 )
> + {
> +
> + unsigned int a_key_len = c->ctl.send_len -
> + strlen("info-get guestinfo.");
> + int rc;
> + struct vmport_state *vs = hd->vmport_data;
> +
> + info_key = (char *)c->send_buf + strlen("info-get guestinfo.");
> + if ( a_key_len <= VMPORT_MAX_KEY_LEN )
> + {
> +
> + rc = vmport_get_guestinfo(hd, vs, info_key, a_key_len,
> + ret_buffer, sizeof(ret_buffer));
> + if ( rc == GUESTINFO_NOTFOUND )
> + ret_msg = "0 No value found";
> + else
> + ret_msg = ret_buffer;
> + }
> + else
> + ret_msg = "0 Key is too long";
> +
> + }
> + else if ( strncmp((char *)c->send_buf, "info-set guestinfo.",
> + strlen("info-set guestinfo.")) == 0 )
> + {
> + char *val;
> + unsigned int rest_len = c->ctl.send_len -
> + strlen("info-set guestinfo.");
> +
> + info_key = (char *)c->send_buf + strlen("info-set guestinfo.");
> + val = strstr(info_key, " ");
> + if ( val )
> + {
> + unsigned int a_key_len = val - info_key;
> + unsigned int a_val_len = rest_len - a_key_len - 1;
> + int rc;
> + struct vmport_state *vs = hd->vmport_data;
> +
> + val++;
> + if ( a_val_len > VMPORT_MAX_VAL_LEN )
> + rc = vmport_set_guestinfo_jumbo(vs, a_key_len,
> + a_val_len,
> + info_key, val);
> + else
> + rc = vmport_set_guestinfo(vs, a_key_len, a_val_len,
> + info_key, val);
> + if ( rc == 0 )
> + ret_msg = "1 ";
> + if ( rc == GUESTINFO_VALTOOLONG )
> + ret_msg = "0 Value too long";
> + if ( rc == GUESTINFO_KEYTOOLONG )
> + ret_msg = "0 Key is too long";
> + if ( rc == GUESTINFO_TOOMANYKEYS )
> + ret_msg = "0 Too many keys";
> +
> +
> + }
> + else
> + ret_msg = "0 Two and exactly two arguments expected";
> + }
> +
> + vmport_send(hd, c, ret_msg);
> + if ( build )
> + {
> + long val = 0;
> + char *p = build;
> +
> + while ( *p )
> + {
> + if ( *p < '0' || *p > '9' )
> + break;
> + val = val * 10 + *p - '0';
> + p++;
> + };
> +
> + hd->params[HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE] = val;
> + hd->params[HVM_PARAM_VMPORT_BUILD_NUMBER_TIME] = now_time;
> + }
> + }
> + else
> + {
> + unsigned int my_bkt = c->ctl.recv_read - 1;
> + vmport_bucket_t *b;
> +
> + if ( my_bkt >= VMPORT_MAX_BKTS )
> + my_bkt = VMPORT_MAX_BKTS - 1;
> + b = &c->recv_bkt[my_bkt];
> + b->ctl.recv_len = 0;
> + }
> + }
> +}
> +
> +static void vmport_process_recv_size(struct hvm_domain *hd, vmport_channel_t *c,
> + struct cpu_user_regs *ur)
> +{
> + vmport_bucket_t *b;
> + vmport_jumbo_bucket_t *jb;
> + int16_t recv_len;
> +
> + if ( c->ctl.jumbo )
> + {
> + jb = &c->jumbo_recv_bkt;
> + recv_len = jb->ctl.recv_len;
> + }
> + else
> + {
> + b = &c->recv_bkt[c->ctl.recv_read];
> + recv_len = b->ctl.recv_len;
> + }
> + if ( recv_len )
> + {
> + set_status(ur, MESSAGE_STATUS_DORECV | MESSAGE_STATUS_SUCCESS);
> + ur->rdx = set_high_bits(ur->rdx, MESSAGE_TYPE_SENDSIZE);
> + ur->rbx = recv_len;
> + }
> + else
> + set_status(ur, MESSAGE_STATUS_SUCCESS);
> +}
> +
> +static void vmport_process_recv_payload(struct hvm_domain *hd,
> + vmport_channel_t *c,
> + struct cpu_user_regs *ur)
> +{
> + vmport_bucket_t *b;
> + vmport_jumbo_bucket_t *jb;
> +
> + if ( c->ctl.jumbo )
> + {
> + jb = &c->jumbo_recv_bkt;
> + ur->rbx = jb->recv_buf[jb->ctl.recv_idx++];
> + }
> + else
> + {
> + b = &c->recv_bkt[c->ctl.recv_read];
> + if ( b->ctl.recv_idx < VMPORT_MAX_RECV_BUF )
> + ur->rbx = b->recv_buf[b->ctl.recv_idx++];
> + else
> + ur->rbx = 0;
> + }
> +
> + set_status(ur, MESSAGE_STATUS_SUCCESS);
> + ur->rdx = set_high_bits(ur->rdx, MESSAGE_TYPE_SENDPAYLOAD);
> +}
> +
> +static void vmport_process_recv_status(struct hvm_domain *hd,
> + vmport_channel_t *c,
> + struct cpu_user_regs *ur)
> +{
> + vmport_bucket_t *b;
> + vmport_jumbo_bucket_t *jb;
> +
> + set_status(ur, MESSAGE_STATUS_SUCCESS);
> +
> + if ( c->ctl.jumbo )
> + {
> + c->ctl.jumbo = 0;
> + /* add debug here */
> + jb = &c->jumbo_recv_bkt;
> + return;
> + }
> +
> + b = &c->recv_bkt[c->ctl.recv_read];
> +
> + c->ctl.recv_read++;
> + if ( c->ctl.recv_read >= VMPORT_MAX_BKTS )
> + c->ctl.recv_read = 0;
> +}
> +
> +static void vmport_process_close(struct hvm_domain *hd, vmport_channel_t *c,
> + struct cpu_user_regs *ur)
> +{
> + /* Return channel to free pool */
> + c->ctl.proto_num = 0;
> + set_status(ur, MESSAGE_STATUS_SUCCESS);
> +}
> +
> +static void vmport_process_packet(struct hvm_domain *hd, vmport_channel_t *c,
> + struct cpu_user_regs *ur, unsigned int sub_cmd,
> + unsigned long now_time)
> +{
> + c->ctl.active_time = now_time;
> +
> + switch ( sub_cmd )
> + {
> + case MESSAGE_TYPE_SENDSIZE:
> + vmport_process_send_size(hd, c, ur);
> + break;
> +
> + case MESSAGE_TYPE_SENDPAYLOAD:
> + vmport_process_send_payload(hd, c, ur, now_time);
> + break;
> +
> + case MESSAGE_TYPE_RECVSIZE:
> + vmport_process_recv_size(hd, c, ur);
> + break;
> +
> + case MESSAGE_TYPE_RECVPAYLOAD:
> + vmport_process_recv_payload(hd, c, ur);
> + break;
> +
> + case MESSAGE_TYPE_RECVSTATUS:
> + vmport_process_recv_status(hd, c, ur);
> + break;
> +
> + case MESSAGE_TYPE_CLOSE:
> + vmport_process_close(hd, c, ur);
> + break;
> +
> + default:
> + ur->rcx = 0;
> + break;
> + }
> +}
> +
> +void vmport_rpc(struct hvm_domain *hd, struct cpu_user_regs *ur)
> +{
> + unsigned int sub_cmd = get_high_bits(ur->rcx);
> + vmport_channel_t *c = NULL;
> + uint16_t msg_id;
> + uint32_t msg_cookie;
> + unsigned long now_time = get_sec();
> + long delta = now_time - hd->vmport_data->ping_time;
> +
> + if ( delta > hd->params[HVM_PARAM_VMPORT_RESET_TIME] )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_PING, "VMware ping. delta=%ld",
> + delta);
> + vmport_ctrl_send(hd, "reset");
> + }
> + spin_lock(&hd->vmport_lock);
> + vmport_sweep(hd, now_time);
> + do {
> + /* Check to see if a new open request is happening... */
> + if ( MESSAGE_TYPE_OPEN == sub_cmd )
> + {
> + c = vmport_new_chan(hd->vmport_data, now_time);
> + if ( NULL == c )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_ERROR,
> + "VMware failed to find a free channel");
> + break;
> + }
> +
> + /* Attach the apropriate protocol the the channel */
> + c->ctl.proto_num = ur->rbx & ~GUESTMSG_FLAG_COOKIE;
> + set_status(ur, MESSAGE_STATUS_SUCCESS);
> + ur->rdx = set_high_bits(ur->rdx, c->ctl.chan_id);
> + ur->rdi = get_low_bits(c->ctl.cookie);
> + ur->rsi = get_high_bits(c->ctl.cookie);
> + if ( c->ctl.proto_num == VMWARE_PROTO_TO_GUEST )
> + vmport_send(hd, c, "reset");
> + break;
> + }
> +
> + msg_id = get_high_bits(ur->rdx);
> + msg_cookie = set_high_bits(ur->rdi, ur->rsi);
> + if ( msg_id >= VMPORT_MAX_CHANS )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware chan id err %d >= %d",
> + msg_id, VMPORT_MAX_CHANS);
> + break;
> + }
> + c = &hd->vmport_data->chans[msg_id];
> + if ( !c->ctl.proto_num )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware chan %d not open",
> + msg_id);
> + break;
> + }
> +
> + /* We check the cookie here since it's possible that the
> + * connection timed out on us and another channel was opened
> + * if this happens, return error and the um tool will
> + * need to reopen the connection
> + */
> + if ( msg_cookie != c->ctl.cookie )
> + {
> + VMPORT_DBG_LOG(VMPORT_LOG_ERROR, "VMware ctl.cookie err %x vs %x",
> + msg_cookie, c->ctl.cookie);
> + break;
> + }
> + vmport_process_packet(hd, c, ur, sub_cmd, now_time);
> + } while ( 0 );
> +
> + if ( NULL == c )
> + set_status(ur, 0);
> +
> + spin_unlock(&hd->vmport_lock);
> +}
> +
> +static int hvm_set_guestinfo(struct vmport_state *vs,
> + struct xen_hvm_vmport_guest_info *a,
> + char *key, char *value)
> +{
> + int idx;
> + int free_idx = -1;
> + int rc = 0;
> +
> + for ( idx = 0; idx < vs->used_guestinfo; idx++ )
> + {
> + if ( !vs->guestinfo[idx] )
> + {
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING,
> + "idx=%d not allocated, but used_guestinfo=%d\n",
> + idx, vs->used_guestinfo);
> +#endif
> + }
> + else if ( (vs->guestinfo[idx]->key_len == a->key_length) &&
> + (memcmp(key,
> + vs->guestinfo[idx]->key_data,
> + vs->guestinfo[idx]->key_len) == 0) )
> + {
> + vs->guestinfo[idx]->val_len = a->value_length;
> + memcpy(vs->guestinfo[idx]->val_data, value, a->value_length);
> + break;
> +
> + }
> + else if ( (vs->guestinfo[idx]->key_len == 0) &&
> + (free_idx == -1) )
> + free_idx = idx;
> + }
> +
> + if ( idx >= vs->used_guestinfo )
> + {
> + if ( free_idx == -1 )
> + rc = -EBUSY;
> + else
> + {
> + vs->guestinfo[free_idx]->key_len = a->key_length;
> + memcpy(vs->guestinfo[free_idx]->key_data, key, a->key_length);
> + vs->guestinfo[free_idx]->val_len = a->value_length;
> + memcpy(vs->guestinfo[free_idx]->val_data, value, a->value_length);
> + }
> + }
> +
> + /* Delete any duplicate entry */
> + if ( rc == 0 )
> + hvm_del_guestinfo_jumbo(vs, key, a->key_length);
> +
> + return rc;
> +}
> +
> +static int hvm_set_guestinfo_jumbo(struct vmport_state *vs,
> + struct xen_hvm_vmport_guest_info *a,
> + char *key, char *value)
> +{
> + int idx;
> + int free_idx = -1;
> + int rc = 0;
> +
> + for ( idx = 0; idx < vs->used_guestinfo_jumbo; idx++ )
> + {
> +
> + if ( !vs->guestinfo_jumbo[idx] )
> + {
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING,
> + "idx=%d not allocated, but used_guestinfo_jumbo=%d\n",
> + idx, vs->used_guestinfo_jumbo);
> +#endif
> + }
> + else if ( (vs->guestinfo_jumbo[idx]->key_len == a->key_length) &&
> + (memcmp(key, vs->guestinfo_jumbo[idx]->key_data,
> + vs->guestinfo_jumbo[idx]->key_len) == 0) )
> + {
> + vs->guestinfo_jumbo[idx]->val_len = a->value_length;
> + memcpy(vs->guestinfo_jumbo[idx]->val_data, value, a->value_length);
> + break;
> +
> + }
> + else if ( (vs->guestinfo_jumbo[idx]->key_len == 0) &&
> + (free_idx == -1) )
> + free_idx = idx;
> + }
> +
> + if ( idx >= vs->used_guestinfo_jumbo )
> + {
> + if ( free_idx == -1 )
> + rc = -EBUSY;
> + else
> + {
> + vs->guestinfo_jumbo[free_idx]->key_len = a->key_length;
> + memcpy(vs->guestinfo_jumbo[free_idx]->key_data,
> + key, a->key_length);
> + vs->guestinfo_jumbo[free_idx]->val_len = a->value_length;
> + memcpy(vs->guestinfo_jumbo[free_idx]->val_data,
> + value, a->value_length);
> + }
> + }
> +
> + /* Delete any duplicate entry */
> + if ( rc == 0 )
> + hvm_del_guestinfo(vs, key, a->key_length);
> +
> + return rc;
> +}
> +
> +static int hvm_get_guestinfo(struct vmport_state *vs,
> + struct xen_hvm_vmport_guest_info *a,
> + char *key, char *value)
> +{
> + int idx;
> + int rc = 0;
> +
> + if ( a->key_length == 0 )
> + {
> + /*
> + * Here we are iterating on getting all guestinfo entries
> + * using index
> + */
> + idx = a->value_length;
> + if ( idx >= vs->used_guestinfo ||
> + !vs->guestinfo[idx] )
> + rc = -ENOENT;
> + else
> + {
> + a->key_length = vs->guestinfo[idx]->key_len;
> + memcpy(a->data, vs->guestinfo[idx]->key_data, a->key_length);
> + a->value_length = vs->guestinfo[idx]->val_len;
> + memcpy(&a->data[a->key_length], vs->guestinfo[idx]->val_data,
> + a->value_length);
> + rc = 0;
> + }
> + }
> + else
> + {
> + for ( idx = 0; idx < vs->used_guestinfo; idx++ )
> + {
> + if ( vs->guestinfo[idx] &&
> + (vs->guestinfo[idx]->key_len == a->key_length) &&
> + (memcmp(key, vs->guestinfo[idx]->key_data,
> + vs->guestinfo[idx]->key_len) == 0) )
> + {
> + a->value_length = vs->guestinfo[idx]->val_len;
> + memcpy(value, vs->guestinfo[idx]->val_data,
> + a->value_length);
> + rc = 0;
> + break;
> + }
> + }
> + if ( idx >= vs->used_guestinfo )
> + rc = -ENOENT;
> + }
> + return rc;
> +}
> +
> +static int hvm_get_guestinfo_jumbo(struct vmport_state *vs,
> + struct xen_hvm_vmport_guest_info *a,
> + char *key, char *value)
> +{
> + int idx, total_entries;
> + int rc = 0;
> +
> + if ( a->key_length == 0 )
> + {
> + /*
> + * Here we are iterating on getting all guestinfo entries
> + * using index
> + */
> + total_entries = vs->used_guestinfo + vs->used_guestinfo_jumbo;
> +
> + /* Input index is in a->value_length */
> + if ( a->value_length >= total_entries )
> + {
> + rc = -ENOENT;
> + return rc;
> + }
> + idx = a->value_length - vs->used_guestinfo;
> + if ( idx >= vs->used_guestinfo_jumbo ||
> + !vs->guestinfo_jumbo[idx] )
> + rc = -ENOENT;
> + else
> + {
> + a->key_length = vs->guestinfo_jumbo[idx]->key_len;
> + memcpy(a->data, vs->guestinfo_jumbo[idx]->key_data, a->key_length);
> + a->value_length = vs->guestinfo_jumbo[idx]->val_len;
> + memcpy(&a->data[a->key_length],
> + vs->guestinfo_jumbo[idx]->val_data, a->value_length);
> + rc = 0;
> + }
> + }
> + else
> + {
> + for ( idx = 0; idx < vs->used_guestinfo_jumbo; idx++ )
> + {
> + if ( vs->guestinfo_jumbo[idx] &&
> + (vs->guestinfo_jumbo[idx]->key_len == a->key_length) &&
> + (memcmp(key, vs->guestinfo_jumbo[idx]->key_data,
> + vs->guestinfo_jumbo[idx]->key_len) == 0) )
> + {
> + a->value_length = vs->guestinfo_jumbo[idx]->val_len;
> + memcpy(value, vs->guestinfo_jumbo[idx]->val_data,
> + a->value_length);
> + rc = 0;
> + break;
> + }
> + }
> + if ( idx >= vs->used_guestinfo_jumbo )
> + rc = -ENOENT;
> + }
> + return rc;
> +}
> +
> +int vmport_rpc_hvmop_precheck(unsigned long op,
> + struct xen_hvm_vmport_guest_info *a)
> +{
> + int new_key_length = a->key_length;
> +
> + if ( new_key_length > strlen("guestinfo.") )
> + {
> + if ( (size_t)new_key_length + (size_t)a->value_length >
> + sizeof(a->data) )
> + return -EINVAL;
> + if ( memcmp(a->data, "guestinfo.", strlen("guestinfo.")) == 0 )
> + new_key_length -= strlen("guestinfo.");
> + if ( new_key_length > VMPORT_MAX_KEY_LEN )
> + {
> + gdprintk(XENLOG_ERR, "bad key len %d\n", new_key_length);
> + return -EINVAL;
> + }
> + if ( a->value_length > VMPORT_MAX_VAL_JUMBO_LEN )
> + {
> + gdprintk(XENLOG_ERR, "bad val len %d\n", a->value_length);
> + return -EINVAL;
> + }
> + }
> + else if ( new_key_length > 0 )
> + {
> + if ( (size_t)new_key_length + (size_t)a->value_length >
> + sizeof(a->data) )
> + return -EINVAL;
> + if ( new_key_length > VMPORT_MAX_KEY_LEN )
> + {
> + gdprintk(XENLOG_ERR, "bad key len %d", new_key_length);
> + return -EINVAL;
> + }
> + if ( a->value_length > VMPORT_MAX_VAL_JUMBO_LEN )
> + {
> + gdprintk(XENLOG_ERR, "bad val len %d\n", a->value_length);
> + return -EINVAL;
> + }
> + }
> + else if ( (new_key_length == 0) && (op == HVMOP_set_vmport_guest_info) )
> + return -EINVAL;
> +
> + return 0;
> +}
> +
> +int vmport_rpc_hvmop_do(struct domain *d, unsigned long op,
> + struct xen_hvm_vmport_guest_info *a)
> +{
> + char *key = NULL;
> + char *value = NULL;
> + struct vmport_state *vs = d->arch.hvm_domain.vmport_data;
> + int i, total_entries;
> + vmport_guestinfo_t *add_slots[5];
> + vmport_guestinfo_jumbo_t *add_slots_jumbo[2];
> + int num_slots = 0, num_slots_jumbo = 0, num_free_slots = 0;
> + int rc = 0;
> +
> + if ( a->key_length > strlen("guestinfo.") )
> + {
> + if ( memcmp(a->data, "guestinfo.", strlen("guestinfo.")) == 0 )
> + {
> + key = &a->data[strlen("guestinfo.")];
> + a->key_length -= strlen("guestinfo.");
> + }
> + else
> + key = &a->data[0];
> + value = key + a->key_length;
> + }
> + else if ( a->key_length > 0 )
> + {
> + key = &a->data[0];
> + value = key + a->key_length;
> + }
> +
> + total_entries = vs->used_guestinfo + vs->used_guestinfo_jumbo;
> +
> + if ( (a->key_length == 0) && (a->value_length >= total_entries) )
> + {
> + /*
> + * When key length is zero, we are interating on
> + * get-guest-info hypercalls to retrieve all guestinfo
> + * entries using index passed in a->value_length
> + */
> + return -E2BIG;
> + }
> +
> + num_free_slots = 0;
> + for ( i = 0; i < vs->used_guestinfo; i++ )
> + {
> + if ( vs->guestinfo[i] &&
> + (vs->guestinfo[i]->key_len == 0) )
> + num_free_slots++;
> + }
> + if ( num_free_slots < 5 )
> + {
> + num_slots = 5 - num_free_slots;
> + if ( vs->used_guestinfo + num_slots > VMPORT_MAX_NUM_KEY )
> + num_slots = VMPORT_MAX_NUM_KEY - vs->used_guestinfo;
> + for ( i = 0; i < num_slots; i++ )
> + add_slots[i] = xzalloc(vmport_guestinfo_t);
> + }
> +
> + num_free_slots = 0;
> + for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
> + {
> + if ( vs->guestinfo_jumbo[i] &&
> + (vs->guestinfo_jumbo[i]->key_len == 0) )
> + num_free_slots++;
> + }
> + if ( num_free_slots < 1 )
> + {
> + num_slots_jumbo = 1 - num_free_slots;
> + if ( vs->used_guestinfo_jumbo + num_slots_jumbo >
> + VMPORT_MAX_NUM_JUMBO_KEY )
> + num_slots_jumbo = VMPORT_MAX_NUM_JUMBO_KEY -
> + vs->used_guestinfo_jumbo;
> + for ( i = 0; i < num_slots_jumbo; i++ )
> + add_slots_jumbo[i] = xzalloc(vmport_guestinfo_jumbo_t);
> + }
> +
> + spin_lock(&d->arch.hvm_domain.vmport_lock);
> +
> + for ( i = 0; i < num_slots; i++ )
> + vs->guestinfo[vs->used_guestinfo + i] = add_slots[i];
> + vs->used_guestinfo += num_slots;
> +
> + for ( i = 0; i < num_slots_jumbo; i++ )
> + vs->guestinfo_jumbo[vs->used_guestinfo_jumbo + i] =
> + add_slots_jumbo[i];
> + vs->used_guestinfo_jumbo += num_slots_jumbo;
> +
> + if ( op == HVMOP_set_vmport_guest_info )
> + {
> + if ( a->value_length > VMPORT_MAX_VAL_LEN )
> + rc = hvm_set_guestinfo_jumbo(vs, a, key, value);
> + else
> + rc = hvm_set_guestinfo(vs, a, key, value);
> + }
> + else
> + {
> + /* Get Guest Info */
> + rc = hvm_get_guestinfo(vs, a, key, value);
> + if ( rc != 0 )
> + rc = hvm_get_guestinfo_jumbo(vs, a, key, value);
> + }
> + spin_unlock(&d->arch.hvm_domain.vmport_lock);
> +
> + return rc;
> +}
> +
> +int vmport_rpc_init(struct domain *d)
> +{
> + struct vmport_state *vs = xzalloc(struct vmport_state);
> + int i;
> +
> + spin_lock_init(&d->arch.hvm_domain.vmport_lock);
> + d->arch.hvm_domain.vmport_data = vs;
> +
> + if ( !vs )
> + return -ENOMEM;
> +
> + /*
> + * Any value is fine here. In fact a random number may better.
> + * It is used to help validate that a both sides are talking
> + * about the same channel.
> + */
> + vs->open_cookie = 435;
> +
> + vs->used_guestinfo = 10;
> + for ( i = 0; i < vs->used_guestinfo; i++ )
> + vs->guestinfo[i] = xzalloc(vmport_guestinfo_t);
> +
> + vs->used_guestinfo_jumbo = 2;
> + for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
> + vs->guestinfo_jumbo[i] = xzalloc(vmport_guestinfo_jumbo_t);
> +
> + vmport_flush(&d->arch.hvm_domain);
> +
> + d->arch.hvm_domain.params[HVM_PARAM_VMPORT_RESET_TIME] = 14;
> +
> + return 0;
> +}
> +
> +void vmport_rpc_deinit(struct domain *d)
> +{
> + struct vmport_state *vs = d->arch.hvm_domain.vmport_data;
> + int i;
> +
> + if ( !vs )
> + return;
> +
> + for ( i = 0; i < vs->used_guestinfo; i++ )
> + xfree(vs->guestinfo[i]);
> + for ( i = 0; i < vs->used_guestinfo_jumbo; i++ )
> + xfree(vs->guestinfo_jumbo[i]);
> + xfree(vs);
> +}
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-set-style: "BSD"
> + * c-basic-offset: 4
> + * tab-width: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/include/asm-x86/hvm/domain.h b/xen/include/asm-x86/hvm/domain.h
> index 93b081b..8fa9090 100644
> --- a/xen/include/asm-x86/hvm/domain.h
> +++ b/xen/include/asm-x86/hvm/domain.h
> @@ -107,6 +107,10 @@ struct hvm_domain {
> /* emulated irq to pirq */
> struct radix_tree_root emuirq_pirq;
>
> + /* VMware special port RPC */
> + spinlock_t vmport_lock;
> + struct vmport_state *vmport_data;
> +
> uint64_t *params;
>
> /* Memory ranges with pinned cache attributes. */
> diff --git a/xen/include/asm-x86/hvm/vmport.h b/xen/include/asm-x86/hvm/vmport.h
> index 74a3b66..a5c70bb 100644
> --- a/xen/include/asm-x86/hvm/vmport.h
> +++ b/xen/include/asm-x86/hvm/vmport.h
> @@ -25,6 +25,15 @@
> #define VMPORT_LOG_VGP_UNKNOWN (1 << 3)
> #define VMPORT_LOG_REALMODE_GP (1 << 4)
>
> +#define VMPORT_LOG_RECV (1 << 8)
> +#define VMPORT_LOG_SEND (1 << 9)
> +#define VMPORT_LOG_SKIP_SEND (1 << 10)
> +#define VMPORT_LOG_ERROR (1 << 11)
> +
> +#define VMPORT_LOG_SWEEP (1 << 12)
> +#define VMPORT_LOG_PING (1 << 13)
> +
> +
> extern unsigned int opt_vmport_debug;
> #define VMPORT_DBG_LOG(level, _f, _a...) \
> do { \
> @@ -42,6 +51,13 @@ int vmport_ioport(int dir, uint32_t port, uint32_t bytes, uint32_t *val);
> int vmport_gp_check(struct cpu_user_regs *regs, struct vcpu *v,
> unsigned long inst_len, unsigned long inst_addr,
> unsigned long ei1, unsigned long ei2);
> +int vmport_rpc_init(struct domain *d);
> +void vmport_rpc_deinit(struct domain *d);
> +void vmport_rpc(struct hvm_domain *hd, struct cpu_user_regs *ur);
> +int vmport_rpc_hvmop_precheck(unsigned long op,
> + struct xen_hvm_vmport_guest_info *a);
> +int vmport_rpc_hvmop_do(struct domain *d, unsigned long op,
> + struct xen_hvm_vmport_guest_info *a);
>
> #endif /* ASM_X86_HVM_VMPORT_H__ */
>
> diff --git a/xen/include/public/arch-x86/hvm/vmporttype.h b/xen/include/public/arch-x86/hvm/vmporttype.h
> new file mode 100644
> index 0000000..98875d2
> --- /dev/null
> +++ b/xen/include/public/arch-x86/hvm/vmporttype.h
> @@ -0,0 +1,118 @@
> +/*
> + * vmporttype.h: HVM VMPORT structure definitions
> + *
> + *
> + * Copyright (C) 2012 Verizon Corporation
> + *
> + * This file is free software; you can redistribute it and/or modify it
> + * under the terms of the GNU General Public License Version 2 (GPLv2)
> + * as published by the Free Software Foundation.
> + *
> + * This file is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> + * General Public License for more details. <http://www.gnu.org/licenses/>.
> + */
> +
> +#ifndef ASM_X86_HVM_VMPORTTYPE_H__
> +#define ASM_X86_HVM_VMPORTTYPE_H__
> +
> +#define VMPORT_MAX_KEY_LEN 30
> +#define VMPORT_MAX_VAL_LEN 128
> +#define VMPORT_MAX_NUM_KEY 64
> +#define VMPORT_MAX_NUM_JUMBO_KEY 4
> +#define VMPORT_MAX_VAL_JUMBO_LEN 4096
> +
> +#define VMPORT_MAX_SEND_BUF ((22 + VMPORT_MAX_KEY_LEN + \
> + VMPORT_MAX_VAL_JUMBO_LEN + 3)/4)
> +#define VMPORT_MAX_RECV_BUF ((2 + VMPORT_MAX_VAL_LEN + 3)/4)
> +#define VMPORT_MAX_RECV_JUMBO_BUF ((2 + VMPORT_MAX_VAL_JUMBO_LEN + 3)/4)
> +#define VMPORT_MAX_CHANS 6
> +#define VMPORT_MAX_BKTS 8
> +
> +#define VMPORT_SAVE_VERSION 0xabcd0001
> +
> +typedef struct
> +{
> + uint8_t key_len;
> + uint8_t val_len;
> + char key_data[VMPORT_MAX_KEY_LEN];
> + char val_data[VMPORT_MAX_VAL_LEN];
> +} vmport_guestinfo_t;
> +
> +typedef struct
> +{
> + uint16_t val_len;
> + uint8_t key_len;
> + char key_data[VMPORT_MAX_KEY_LEN];
> + char val_data[VMPORT_MAX_VAL_JUMBO_LEN];
> +} vmport_guestinfo_jumbo_t;
> +
> +typedef struct __attribute__((packed))
> +{
> + uint16_t recv_len;
> + uint16_t recv_idx;
> +}
> +vmport_bucket_control_t;
> +
> +typedef struct __attribute__((packed))
> +{
> + vmport_bucket_control_t ctl;
> + uint32_t recv_buf[VMPORT_MAX_RECV_BUF + 1];
> +}
> +vmport_bucket_t;
> +
> +typedef struct __attribute__((packed))
> +{
> + vmport_bucket_control_t ctl;
> + uint32_t recv_buf[VMPORT_MAX_RECV_JUMBO_BUF + 1];
> +}
> +vmport_jumbo_bucket_t;
> +
> +typedef struct __attribute__((packed))
> +{
> + uint64_t active_time;
> + uint32_t chan_id;
> + uint32_t cookie;
> + uint32_t proto_num;
> + uint16_t send_len;
> + uint16_t send_idx;
> + uint8_t jumbo;
> + uint8_t recv_read;
> + uint8_t recv_write;
> + uint8_t recv_chan_pad[1];
> +}
> +vmport_channel_control_t;
> +
> +typedef struct __attribute__((packed))
> +{
> + vmport_channel_control_t ctl;
> + vmport_bucket_t recv_bkt[VMPORT_MAX_BKTS];
> + vmport_jumbo_bucket_t jumbo_recv_bkt;
> + uint32_t send_buf[VMPORT_MAX_SEND_BUF + 1];
> +}
> +vmport_channel_t;
> +
> +struct vmport_state
> +{
> + uint64_t ping_time;
> + uint32_t open_cookie;
> + uint32_t used_guestinfo;
> + uint32_t used_guestinfo_jumbo;
> + uint8_t max_chans;
> + uint8_t state_pad[3];
> + vmport_channel_t chans[VMPORT_MAX_CHANS];
> + vmport_guestinfo_t *guestinfo[VMPORT_MAX_NUM_KEY];
> + vmport_guestinfo_jumbo_t *guestinfo_jumbo[VMPORT_MAX_NUM_JUMBO_KEY];
> +};
> +
> +#endif /* ASM_X86_HVM_VMPORTTYPE_H__ */
> +
> +/*
> + * Local variables:
> + * mode: C
> + * c-file-style: "BSD"
> + * c-basic-offset: 4
> + * indent-tabs-mode: nil
> + * End:
> + */
> diff --git a/xen/include/public/hvm/hvm_op.h b/xen/include/public/hvm/hvm_op.h
> index eeb0a60..8e1e072 100644
> --- a/xen/include/public/hvm/hvm_op.h
> +++ b/xen/include/public/hvm/hvm_op.h
> @@ -369,6 +369,24 @@ DEFINE_XEN_GUEST_HANDLE(xen_hvm_set_ioreq_server_state_t);
>
> #endif /* defined(__XEN__) || defined(__XEN_TOOLS__) */
>
> +/* Get/set vmport subcommands */
> +#define HVMOP_get_vmport_guest_info 23
> +#define HVMOP_set_vmport_guest_info 24
> +#define VMPORT_GUEST_INFO_KEY_MAX 40
> +#define VMPORT_GUEST_INFO_VAL_MAX 4096
> +struct xen_hvm_vmport_guest_info {
> + /* Domain to be accessed */
> + domid_t domid;
> + /* key length */
> + uint16_t key_length;
> + /* value length */
> + uint16_t value_length;
> + /* key and value data */
> + char data[VMPORT_GUEST_INFO_KEY_MAX + VMPORT_GUEST_INFO_VAL_MAX];
> +};
> +typedef struct xen_hvm_vmport_guest_info xen_hvm_vmport_guest_info_t;
> +DEFINE_XEN_GUEST_HANDLE(xen_hvm_vmport_guest_info_t);
> +
> #endif /* __XEN_PUBLIC_HVM_HVM_OP_H__ */
>
> /*
> diff --git a/xen/include/public/hvm/params.h b/xen/include/public/hvm/params.h
> index dee6d68..722e30b 100644
> --- a/xen/include/public/hvm/params.h
> +++ b/xen/include/public/hvm/params.h
> @@ -153,7 +153,10 @@
>
> /* Params for VMware */
> #define HVM_PARAM_VMWARE_HW 35
> +#define HVM_PARAM_VMPORT_BUILD_NUMBER_TIME 36
> +#define HVM_PARAM_VMPORT_BUILD_NUMBER_VALUE 37
> +#define HVM_PARAM_VMPORT_RESET_TIME 38
>
> -#define HVM_NR_PARAMS 36
> +#define HVM_NR_PARAMS 39
>
> #endif /* __XEN_PUBLIC_HVM_PARAMS_H__ */
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 11/16] Add live migration of VMware's hyper-call RPC
2014-09-11 18:36 ` [PATCH v4 11/16] Add live migration of VMware's hyper-call RPC Don Slutz
@ 2014-09-12 13:54 ` Boris Ostrovsky
2014-09-16 15:48 ` Don Slutz
0 siblings, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-12 13:54 UTC (permalink / raw)
To: Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/11/2014 02:36 PM, Don Slutz wrote:
> The VMware's hyper-call state is included in live migration and
> save/restore. Because the max size of the VMware guestinfo is
> large, then data is compressed and expanded in the
> vmport_save_domain_ctxt and vmport_load_domain_ctxt.
>
> Signed-off-by: Don Slutz <dslutz@verizon.com>
> ---
> xen/arch/x86/hvm/vmware/vmport_rpc.c | 302 ++++++++++++++++++++++++++++++++-
> xen/include/public/arch-x86/hvm/save.h | 39 ++++-
> 2 files changed, 339 insertions(+), 2 deletions(-)
>
> diff --git a/xen/arch/x86/hvm/vmware/vmport_rpc.c b/xen/arch/x86/hvm/vmware/vmport_rpc.c
> index 695e58f..e25ed09 100644
> --- a/xen/arch/x86/hvm/vmware/vmport_rpc.c
> +++ b/xen/arch/x86/hvm/vmware/vmport_rpc.c
> @@ -39,7 +39,8 @@
> #include <asm/hvm/vmport.h>
> #include <asm/hvm/trace.h>
>
> -#include "public/arch-x86/hvm/vmporttype.h"
> +#include <public/hvm/save.h>
> +#include <public/arch-x86/hvm/save.h>
>
> #include "backdoor_def.h"
> #include "guest_msg_def.h"
> @@ -1270,6 +1271,305 @@ void vmport_rpc_deinit(struct domain *d)
> xfree(vs);
> }
>
> +/* save and restore functions */
> +
> +static int vmport_save_domain_ctxt(struct domain *d, hvm_domain_context_t *h)
> +{
> + struct hvm_vmport_context *ctxt;
> + struct hvm_save_descriptor *desc;
> + struct hvm_domain *hd = &d->arch.hvm_domain;
> + struct vmport_state *vs = hd->vmport_data;
> + char *p;
> + unsigned int guestinfo_size = 0;
> + unsigned int used_guestinfo = 0;
> + unsigned int used_guestinfo_jumbo = 0;
> + unsigned int chans_size;
> + unsigned int i;
> +
> + /* Customized handling for entry since our entry is of variable length */
> + desc = (struct hvm_save_descriptor *)&h->data[h->cur];
> + if ( _hvm_init_entry(h, HVM_SAVE_CODE(VMPORT), 0,
> + HVM_SAVE_LENGTH(VMPORT)) )
> + return 1;
> + ctxt = (struct hvm_vmport_context *)&h->data[h->cur];
> +
> + spin_lock(&hd->vmport_lock);
> +
> + ctxt->version = VMPORT_SAVE_VERSION;
> + ctxt->ping_time = vs->ping_time;
> + ctxt->open_cookie = vs->open_cookie;
> + ctxt->used_guestinfo = vs->used_guestinfo;
> + ctxt->used_guestinfo_jumbo = vs->used_guestinfo_jumbo;
> +
> + p = ctxt->u.packed.packed_data;
> +
> + for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
> + {
> + unsigned int j;
> + unsigned int buf_max;
> +
> + ctxt->u.packed.chan_ctl[i].chan = vs->chans[i].ctl;
> + buf_max = vs->chans[i].ctl.send_len;
> + if ( buf_max > VMPORT_MAX_SEND_BUF * 4 )
> + buf_max = VMPORT_MAX_SEND_BUF * 4;
> + memcpy(p, vs->chans[i].send_buf, buf_max);
> + p += buf_max;
> + for ( j = 0; j < VMPORT_MAX_BKTS; j++ )
> + {
> + ctxt->u.packed.chan_ctl[i].recv[j] = vs->chans[i].recv_bkt[j].ctl;
> + buf_max = vs->chans[i].recv_bkt[j].ctl.recv_len;
> + if ( buf_max > VMPORT_MAX_RECV_BUF * 4 )
> + buf_max = VMPORT_MAX_RECV_BUF * 4;
> + memcpy(p, vs->chans[i].recv_bkt[j].recv_buf, buf_max);
> + p += buf_max;
> + }
> + ctxt->u.packed.chan_ctl[i].jumbo = vs->chans[i].jumbo_recv_bkt.ctl;
> + buf_max = vs->chans[i].jumbo_recv_bkt.ctl.recv_len;
> + if ( buf_max > VMPORT_MAX_RECV_JUMBO_BUF * 4 )
> + buf_max = VMPORT_MAX_RECV_JUMBO_BUF * 4;
> + memcpy(p, vs->chans[i].jumbo_recv_bkt.recv_buf, buf_max);
> + p += buf_max;
> + }
> +
> + chans_size = p - ctxt->u.packed.packed_data;
> +
> + for ( i = 0; i < ctxt->used_guestinfo; i++ )
> + {
> + vmport_guestinfo_t *vg = vs->guestinfo[i];
> +
> + if ( vg && vg->key_len )
> + {
> + guestinfo_size += sizeof(vg->key_len) + sizeof(vg->val_len) +
> + vg->key_len + vg->val_len;
> + used_guestinfo++;
> + ASSERT(sizeof(vg->key_len) == 1);
> + *p++ = (char) vg->key_len;
> + ASSERT(sizeof(vg->val_len) == 1);
> + *p++ = (char) vg->val_len;
> + if ( vg->key_len )
You ASSERTed that vg->key_len is 1 so you may not need the 'if'. (And in
NDEBUG version the worst that could happen is you do memcpy(,,0), which
is safe).
> + {
> + memcpy(p, vg->key_data, vg->key_len);
> + p += vg->key_len;
> + if ( vg->val_len )
Same here.
> + {
> + memcpy(p, vg->val_data, vg->val_len);
> + p += vg->val_len;
> + }
> + }
> + }
> + }
> + ctxt->used_guestinfo = used_guestinfo;
> +
> + for ( i = 0; i < ctxt->used_guestinfo_jumbo; i++ )
> + {
> + vmport_guestinfo_jumbo_t *vgj =
> + vs->guestinfo_jumbo[i];
> + if ( vgj && vgj->key_len )
> + {
> + guestinfo_size += sizeof(vgj->key_len) + sizeof(vgj->val_len) +
> + vgj->key_len + vgj->val_len;
> + used_guestinfo_jumbo++;
> + ASSERT(sizeof(vgj->key_len) == 1);
> + *p++ = (char) vgj->key_len;
> + memcpy(p, &vgj->val_len, sizeof(vgj->val_len));
> + p += sizeof(vgj->val_len);
> + if ( vgj->key_len )
And here.
And do you need ASSERT(vgj->val_len ==1) as well? Also, In
vmport_load_domain_ctxt() you don't seem to be making this assertions. I
don't know whether this is on purpose.
> + {
> + memcpy(p, vgj->key_data, vgj->key_len);
> + p += vgj->key_len;
> + if ( vgj->val_len )
> + {
> + memcpy(p, vgj->val_data, vgj->val_len);
> + p += vgj->val_len;
> + }
> + }
> + }
> + }
> + ctxt->used_guestinfo_jumbo = used_guestinfo_jumbo;
> +
> + ctxt->used_guestsize = guestinfo_size;
> +
> + spin_unlock(&hd->vmport_lock);
> +
> +#ifndef NDEBUG
> + gdprintk(XENLOG_WARNING, "chans_size=%d guestinfo_size=%d, used=%ld\n",
> + chans_size, guestinfo_size,
> + p - ctxt->u.packed.packed_data);
> +#endif
> + ASSERT(p - ctxt->u.packed.packed_data == chans_size + guestinfo_size);
> + ASSERT(desc->length >= p - (char *)ctxt);
> + desc->length = p - (char *)ctxt; /* Fixup length to be right */
> + h->cur += desc->length; /* Do _hvm_write_entry */
> + ASSERT(guestinfo_size < desc->length);
> +
> + return 0;
> +}
> +
> +static int vmport_load_domain_ctxt(struct domain *d, hvm_domain_context_t *h)
> +{
> + struct hvm_vmport_context *ctxt;
> + struct hvm_save_descriptor *desc;
> + struct hvm_domain *hd = &d->arch.hvm_domain;
> + struct vmport_state *vs = hd->vmport_data;
> + unsigned int i;
> + uint8_t key_len;
> + uint16_t val_len;
> + char *p;
> + vmport_guestinfo_t *vg;
> + vmport_guestinfo_jumbo_t *vgj;
> + unsigned int loop_cnt;
> + unsigned int guestinfo_size;
> + unsigned int used_guestinfo;
> + unsigned int used_guestinfo_jumbo;
> +
> + if ( !vs )
> + return -ENOMEM;
> +
> + /* Customized checking for entry since our entry is of variable length */
> + desc = (struct hvm_save_descriptor *)&h->data[h->cur];
> + if ( sizeof(*desc) > h->size - h->cur )
> + {
> + printk(XENLOG_G_WARNING
> + "HVM%d restore: not enough data left to read descriptor"
> + "for type %lu\n", d->domain_id,
> + HVM_SAVE_CODE(VMPORT));
> + return -1;
Since in some cases you are returning proper error codes you should
return them everywhere.
-boris
> + }
> + if ( desc->length + sizeof(*desc) > h->size - h->cur )
> + {
> + printk(XENLOG_G_WARNING
> + "HVM%d restore: not enough data left to read %u bytes "
> + "for type %lu\n", d->domain_id, desc->length,
> + HVM_SAVE_CODE(VMPORT));
> + return -1;
> + }
> + if ( HVM_SAVE_CODE(VMPORT) != desc->typecode ||
> + (desc->length > HVM_SAVE_LENGTH(VMPORT)) )
> + {
> + printk(XENLOG_G_WARNING
> + "HVM%d restore mismatch: expected type %lu with max length %lu, "
> + "saw type %u length %u\n", d->domain_id, HVM_SAVE_CODE(VMPORT),
> + HVM_SAVE_LENGTH(VMPORT),
> + desc->typecode, desc->length);
> + return -1;
> + }
> + h->cur += sizeof(*desc);
> + /* Checking finished */
> +
> + ctxt = (struct hvm_vmport_context *)&h->data[h->cur];
> + h->cur += desc->length;
> +
> + if ( ctxt->version != VMPORT_SAVE_VERSION )
> + return -EINVAL;
> +
> + spin_lock(&hd->vmport_lock);
> +
> + vs->ping_time = ctxt->ping_time;
> + vs->open_cookie = ctxt->open_cookie;
> + vs->used_guestinfo = ctxt->used_guestinfo;
> + vs->used_guestinfo_jumbo = ctxt->used_guestinfo_jumbo;
> + guestinfo_size = ctxt->used_guestsize;
> + used_guestinfo = ctxt->used_guestinfo;
> + used_guestinfo_jumbo = ctxt->used_guestinfo_jumbo;
> +
> + p = ctxt->u.packed.packed_data;
> +
> + for ( i = 0; i < VMPORT_MAX_CHANS; i++ )
> + {
> + unsigned int j;
> +
> + vs->chans[i].ctl = ctxt->u.packed.chan_ctl[i].chan;
> + memcpy(vs->chans[i].send_buf, p, vs->chans[i].ctl.send_len);
> + p += vs->chans[i].ctl.send_len;
> + for ( j = 0; j < VMPORT_MAX_BKTS; j++ )
> + {
> + vs->chans[i].recv_bkt[j].ctl = ctxt->u.packed.chan_ctl[i].recv[j];
> + memcpy(vs->chans[i].recv_bkt[j].recv_buf, p,
> + vs->chans[i].recv_bkt[j].ctl.recv_len);
> + p += vs->chans[i].recv_bkt[j].ctl.recv_len;
> + }
> + vs->chans[i].jumbo_recv_bkt.ctl = ctxt->u.packed.chan_ctl[i].jumbo;
> + memcpy(vs->chans[i].jumbo_recv_bkt.recv_buf, p,
> + vs->chans[i].jumbo_recv_bkt.ctl.recv_len);
> + p += vs->chans[i].jumbo_recv_bkt.ctl.recv_len;
> + }
> +
> +
> + /* keep at least 10 total and 5 empty entries */
> + loop_cnt = (vs->used_guestinfo + 5) > 10 ?
> + (vs->used_guestinfo + 5) : 10;
> + for ( i = 0; i < loop_cnt; i++ )
> + {
> + if ( !vs->guestinfo[i] )
> + {
> + vs->guestinfo[i] = xzalloc(vmport_guestinfo_t);
> + }
> + if ( i < vs->used_guestinfo
> + && guestinfo_size > 0 )
> + {
> + key_len = (uint8_t)*p++;
> + val_len = (uint8_t)*p++;
> + guestinfo_size -= 2;
> + if ( guestinfo_size >= key_len + val_len )
> + {
> + vg = vs->guestinfo[i];
> + if ( key_len )
> + {
> + vg->key_len = key_len;
> + vg->val_len = val_len;
> + memcpy(vg->key_data, p, key_len);
> + p += key_len;
> + memcpy(vg->val_data, p, val_len);
> + p += val_len;
> + guestinfo_size -= key_len + val_len;
> + }
> + }
> + }
> + }
> + vs->used_guestinfo = loop_cnt;
> +
> + /* keep at least 2 total and 1 empty entries */
> + loop_cnt = (vs->used_guestinfo_jumbo + 1) > 2 ?
> + (vs->used_guestinfo_jumbo + 1) : 2;
> + for ( i = 0; i < loop_cnt; i++ )
> + {
> + if ( !vs->guestinfo_jumbo[i] )
> + {
> + vs->guestinfo_jumbo[i] = xzalloc(vmport_guestinfo_jumbo_t);
> + }
> + if ( i < vs->used_guestinfo_jumbo
> + && guestinfo_size > 0 )
> + {
> + key_len = (uint8_t)*p++;
> + memcpy(&val_len, p, 2);
> + p += 2;
> + guestinfo_size -= 3;
> + if ( guestinfo_size >= key_len + val_len )
> + {
> + vgj = vs->guestinfo_jumbo[i];
> + if ( key_len )
> + {
> + vgj->key_len = key_len;
> + vgj->val_len = val_len;
> + memcpy(vgj->key_data, p, key_len);
> + p += key_len;
> + memcpy(vgj->val_data, p, val_len);
> + p += val_len;
> + guestinfo_size -= key_len + val_len;
> + }
> + }
> + }
> + }
> + vs->used_guestinfo_jumbo = loop_cnt;
> +
> + spin_unlock(&hd->vmport_lock);
> +
> + return 0;
> +}
> +
> +HVM_REGISTER_SAVE_RESTORE(VMPORT, vmport_save_domain_ctxt,
> + vmport_load_domain_ctxt, 1, HVMSR_PER_DOM);
> +
> /*
> * Local variables:
> * mode: C
> diff --git a/xen/include/public/arch-x86/hvm/save.h b/xen/include/public/arch-x86/hvm/save.h
> index 16d85a3..7cdcb8f 100644
> --- a/xen/include/public/arch-x86/hvm/save.h
> +++ b/xen/include/public/arch-x86/hvm/save.h
> @@ -26,6 +26,8 @@
> #ifndef __XEN_PUBLIC_HVM_SAVE_X86_H__
> #define __XEN_PUBLIC_HVM_SAVE_X86_H__
>
> +#include "vmporttype.h"
> +
> /*
> * Save/restore header: general info about the save file.
> */
> @@ -610,9 +612,44 @@ struct hvm_msr {
>
> #define CPU_MSR_CODE 20
>
> +/*
> + * VMware context.
> + */
> +struct hvm_vmport_context {
> + uint32_t version;
> + uint32_t used_guestsize;
> + uint64_t ping_time;
> + uint32_t open_cookie;
> + uint32_t used_guestinfo;
> + uint32_t used_guestinfo_jumbo;
> + union {
> + struct {
> + vmport_channel_t chans[VMPORT_MAX_CHANS];
> + vmport_guestinfo_t guestinfo[VMPORT_MAX_NUM_KEY];
> + vmport_guestinfo_jumbo_t guestinfo_jumbo[VMPORT_MAX_NUM_JUMBO_KEY];
> + } full;
> + struct {
> + struct {
> + vmport_channel_control_t chan;
> + vmport_bucket_control_t recv[VMPORT_MAX_BKTS];
> + vmport_bucket_control_t jumbo;
> + } chan_ctl[VMPORT_MAX_CHANS];
> +#if defined(__STDC_VERSION__) && __STDC_VERSION__ >= 199901L
> + char packed_data[];
> +#elif defined(__GNUC__)
> + char packed_data[0];
> +#else
> + char packed_data[1 /* Variable length */];
> +#endif
> + } packed;
> + } u;
> +};
> +
> +DECLARE_HVM_SAVE_TYPE(VMPORT, 21, struct hvm_vmport_context);
> +
> /*
> * Largest type-code in use
> */
> -#define HVM_SAVE_CODE_MAX 20
> +#define HVM_SAVE_CODE_MAX 21
>
> #endif /* __XEN_PUBLIC_HVM_SAVE_X86_H__ */
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc
2014-09-12 13:37 ` Boris Ostrovsky
@ 2014-09-12 14:27 ` Jan Beulich
2014-09-16 12:38 ` Slutz, Donald Christopher
2014-09-16 13:17 ` Slutz, Donald Christopher
1 sibling, 1 reply; 51+ messages in thread
From: Jan Beulich @ 2014-09-12 14:27 UTC (permalink / raw)
To: xen-devel, Boris Ostrovsky, Don Slutz
Cc: Jun Nakajima, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
Tim Deegan, Eddie Dong, Aravind Gopalakrishnan,
Suravee Suthikulpanit
>>> On 12.09.14 at 15:37, <boris.ostrovsky@oracle.com> wrote:
> On 09/11/2014 02:36 PM, Don Slutz wrote:
>> +static inline uint16_t get_low_bits(uint32_t bits)
>> +{
>> + return bits & 0xffff;
>> +}
>> +
>> +static inline uint16_t get_high_bits(uint32_t bits)
>> +{
>> + return bits >> 16;
>> +}
>> +
>> +static inline uint32_t set_high_bits(uint32_t b, uint32_t val)
>> +{
>> + return (val << 16) | get_low_bits(b);
>> +}
The names of all three functions (at least one of which I think I saw
in a reply to an earlier patch - code duplication?) are bogus: How
many high or low bits (and for the "high" case out of how many) are
we talking about here?
>> +static inline void set_status(struct cpu_user_regs *ur, uint16_t val)
>> +{
>> + /* VMware defines this to be only 32 bits */
>> + ur->rcx = (val << 16) | (ur->rcx & 0xffff);
>
> Are you shifting val out of its size? Or will compiler automatically
> promote this to uint32_t? (It does for me but I don't know if this is
> something we can rely on).
All operands more narrow than int get promoted to int.
Jan
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-12 9:49 ` Jan Beulich
@ 2014-09-12 17:46 ` Don Slutz
2014-09-15 7:42 ` Jan Beulich
0 siblings, 1 reply; 51+ messages in thread
From: Don Slutz @ 2014-09-12 17:46 UTC (permalink / raw)
To: Jan Beulich, Andrew Cooper, xen-devel, Don Slutz
Cc: Jun Nakajima, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Eddie Dong, Ian Jackson,
Tim Deegan, Aravind Gopalakrishnan, Suravee Suthikulpanit,
Boris Ostrovsky
On 09/12/14 05:49, Jan Beulich wrote:
>>>> On 11.09.14 at 21:49, <andrew.cooper3@citrix.com> wrote:
>>>>
>>> + case 0x10:
>>> + /* (Virtual) TSC frequency in kHz. */
>>> + *eax = d->arch.tsc_khz;
>>> + /* (Virtual) Bus (local apic timer) frequency in kHz. */
>>> + *ebx = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
>> At least 1 pair of brackets please, especially as the placement of
>> brackets affects the result of this particular calculation.
> Or simply eliminate one of the divisions using
> "1000000ull / APIC_BUS_CYCLE_NS".
I am totally confused. I am happy to go with Jan's version.
The confusion is that I get the same answer all the ways I try. This
includes
when APIC_BUS_CYCLE_NS is other values then 10.
Some facts:
1) 1000000000 takes only 30bits, I am asking it to be 64.
2) I am only making it smaller in all cases.
I have tested this on CentOS release 6.5 (Final) both 32 and 64 and
it all says they are all the same.
Here is my testing:
build1:~/tmp/zz-div>p /etc/redhat-release
CentOS release 6.5 (Final)
build1:~/tmp/zz-div>uname -a
Linux build1 2.6.32-431.el6.i686 #1 SMP Fri Nov 22 00:26:36 UTC 2013
i686 i686 i386 GNU/Linux
build1:~/tmp/zz-div>cat zz-div.c
#include <stdio.h>
#include <stdint.h>
int main(void)
{
uint32_t ebx1, ebx2, ebx3, ebx4, ebx5;
uint32_t start = 1000000000ull;
uint32_t APIC_BUS_CYCLE_NS;
printf("start=0x%x, %d\n\n", start, start);
for ( APIC_BUS_CYCLE_NS = 1; APIC_BUS_CYCLE_NS < 10000;
APIC_BUS_CYCLE_NS++ )
{
ebx1 = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
ebx2 = (1000000000ull / APIC_BUS_CYCLE_NS) / 1000ull;
ebx3 = 1000000000ull / (APIC_BUS_CYCLE_NS * 1000ull);
ebx4 = 1000000000ull / 1000ull / APIC_BUS_CYCLE_NS;
ebx5 = 1000000ull / APIC_BUS_CYCLE_NS;
if ( ebx1 != ebx2 ||
ebx1 != ebx3 ||
ebx1 != ebx4 ||
ebx1 != ebx5 )
{
printf("APIC_BUS_CYCLE_NS=%d:\n", APIC_BUS_CYCLE_NS);
printf(" ebx1=0x%x, %d\n", ebx1, ebx1);
printf(" ebx2=0x%x, %d\n", ebx2, ebx2);
printf(" ebx3=0x%x, %d\n", ebx3, ebx3);
printf(" ebx4=0x%x, %d\n", ebx4, ebx4);
printf(" ebx5=0x%x, %d\n", ebx5, ebx5);
}
}
return 0;
}
/*
* Local variables:
* mode: C
* c-file-style: "BSD"
* c-basic-offset: 4
* indent-tabs-mode: nil
* End:
*/
build1:~/tmp/zz-div>gcc -g -o zz-div zz-div.c
build1:~/tmp/zz-div>./zz-div
start=0x3b9aca00, 1000000000
build1:~/tmp/zz-div>
-Don Slutz
>
>
> Jan
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-12 12:37 ` Boris Ostrovsky
@ 2014-09-12 17:50 ` Don Slutz
0 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-12 17:50 UTC (permalink / raw)
To: Boris Ostrovsky, Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/12/14 08:37, Boris Ostrovsky wrote:
> On 09/11/2014 02:36 PM, Don Slutz wrote:
>> diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
>> index 10fc2ca..f353f42 100644
>> --- a/xen/arch/x86/traps.c
>> +++ b/xen/arch/x86/traps.c
>> @@ -685,8 +685,12 @@ int cpuid_hypervisor_leaves( uint32_t idx,
>> uint32_t sub_idx,
>> uint32_t *eax, uint32_t *ebx, uint32_t *ecx,
>> uint32_t *edx)
>> {
>> struct domain *d = current->domain;
>> - /* Optionally shift out of the way of Viridian architectural
>> leaves. */
>> - uint32_t base = is_viridian_domain(d) ? 0x40000100 : 0x40000000;
>> + /*
>> + * Optionally shift out of the way of Viridian or VMware
>> + * architectural leaves.
>> + */
>> + uint32_t base = is_viridian_domain(d) | is_vmware_domain(d) ?
>
> Given how is_viridian and is_vmware are defined I think '||' is more
> appropriate.
>
I plan to convert to "||".
-Don Slutz
> -boris
>
>>
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-11 19:49 ` Andrew Cooper
2014-09-12 9:49 ` Jan Beulich
@ 2014-09-12 21:26 ` Don Slutz
1 sibling, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-12 21:26 UTC (permalink / raw)
To: Andrew Cooper, Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Boris Ostrovsky,
Suravee Suthikulpanit
On 09/11/14 15:49, Andrew Cooper wrote:
> On 11/09/2014 19:36, Don Slutz wrote:
>> This is done by adding HVM_PARAM_VMWARE_HW. It is set to the VMware
>> virtual hardware version.
>>
>> Currently 0, 3-4, 6-11 are good values. However the
>> code only checks for == 0 or != 0.
>>
>> If non-zero then
>> Return VMware's cpuid leaves.
>>
>> The support of hypervisor cpuid leaves has not been agreed to.
>>
>> MicroSoft Hyper-V (AKA viridian) currently must be at 0x40000000.
>>
>> VMware currently must be at 0x40000000.
>>
>> KVM currently must be at 0x40000000 (from Seabios).
>>
>> Seabios will find xen at 0x40000000, 0x40000100, 0x40000200 ..
>> 0x40010000.
>>
>> http://download.microsoft.com/download/F/B/0/FB0D01A3-8E3A-4F5F-AA59-08C8026D3B8A/requirements-for-implementing-microsoft-hypervisor-interface.docx
>>
>> http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1009458
>>
>> http://lwn.net/Articles/301888/
>> Attempted to get this cleaned up.
>>
>> So based on this, I picked the order:
>>
>> Xen at 0x40000000 or
>> Viridian or VMware at 0x40000000 and Xen at 0x40000100
>>
>> If both Viridian and VMware selected, report an error.
>>
>> Signed-off-by: Don Slutz <dslutz@verizon.com>
>> ---
>> xen/arch/x86/hvm/Makefile | 3 +-
>> xen/arch/x86/hvm/hvm.c | 32 +++++++++++++++++++
>> xen/arch/x86/hvm/vmware/Makefile | 1 +
>> xen/arch/x86/hvm/vmware/cpuid.c | 69 ++++++++++++++++++++++++++++++++++++++++
>> xen/arch/x86/traps.c | 8 +++--
>> xen/include/asm-x86/hvm/hvm.h | 3 ++
>> xen/include/asm-x86/hvm/vmware.h | 31 ++++++++++++++++++
>> xen/include/public/hvm/params.h | 5 ++-
>> 8 files changed, 148 insertions(+), 4 deletions(-)
>> create mode 100644 xen/arch/x86/hvm/vmware/Makefile
>> create mode 100644 xen/arch/x86/hvm/vmware/cpuid.c
>> create mode 100644 xen/include/asm-x86/hvm/vmware.h
>>
>> diff --git a/xen/arch/x86/hvm/Makefile b/xen/arch/x86/hvm/Makefile
>> index eea5555..77598a6 100644
>> --- a/xen/arch/x86/hvm/Makefile
>> +++ b/xen/arch/x86/hvm/Makefile
>> @@ -1,5 +1,6 @@
>> subdir-y += svm
>> subdir-y += vmx
>> +subdir-y += vmware
>>
>> obj-y += asid.o
>> obj-y += emulate.o
>> @@ -22,4 +23,4 @@ obj-y += vlapic.o
>> obj-y += vmsi.o
>> obj-y += vpic.o
>> obj-y += vpt.o
>> -obj-y += vpmu.o
>> \ No newline at end of file
>> +obj-y += vpmu.o
> This hunk is unrelated, but is perhaps something better fixed. A
> passing note in the commit message perhaps?
>
Sure.
[snip -- different thread]
> +#include <xen/sched.h>
> +
> +#include <asm/hvm/hvm.h>
> +#include <asm/hvm/vmware.h>
> +
> +int cpuid_vmware_leaves(uint32_t idx, uint32_t *eax, uint32_t *ebx,
> + uint32_t *ecx, uint32_t *edx)
> +{
> + struct domain *d = current->domain;
> +
> + if ( !is_vmware_domain(d) )
> + return 0;
> +
> + switch ( idx - 0x40000000 )
> + {
> + case 0x0:
> Do these leaves have semantic names from a spec, which could live in an
> appropriate header file?
Not sure.
http://lwn.net/Articles/301888/
Which is from Oct 2008 -- Looks like VMware went with this and no one
else.
I can only find list postings that refer the the statement:
"
VMware hardware version 7 defines some of these cpuid levels, below is a
brief
description about those. These levels can be implemented by other
hypervisors
too so that Linux has a standard way of communicating to any hypervisor.
Leaf 0x40000000, Hypervisor CPUID information
# EAX: The maximum input value for hypervisor CPUID info (0x40000010).
# EBX, ECX, EDX: Hypervisor vendor ID signature. E.g. "VMwareVMware"
Leaf 0x40000010, Timing information.
# EAX: (Virtual) TSC frequency in kHz.
# EBX: (Virtual) Bus (local apic timer) frequency in kHz.
# ECX, EDX: RESERVED
"
There are a few about 0x40000010 as HYPERVISOR_TIMING_LEAF
or HYPERVISOR_VMWARE_TIMING_LEAF.
All from late Sep 2008 (https://lkml.org/lkml/2008/9/29/45)
But I also found that I should be checking for >= 7:
The hypervisor present bit and hypervisor information leaf are only
defined for products based on VMware hardware version 7.
So I will be changing this.
>> + case 0x10:
>> + /* (Virtual) TSC frequency in kHz. */
>> + *eax = d->arch.tsc_khz;
>> + /* (Virtual) Bus (local apic timer) frequency in kHz. */
>> + *ebx = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
> At least 1 pair of brackets please, especially as the placement of
> brackets affects the result of this particular calculation.
See other thread.
>> diff --git a/xen/arch/x86/traps.c b/xen/arch/x86/traps.c
>> index 10fc2ca..f353f42 100644
>> --- a/xen/arch/x86/traps.c
>> +++ b/xen/arch/x86/traps.c
>> @@ -685,8 +685,12 @@ int cpuid_hypervisor_leaves( uint32_t idx, uint32_t sub_idx,
>> uint32_t *eax, uint32_t *ebx, uint32_t *ecx, uint32_t *edx)
>> {
>> struct domain *d = current->domain;
>> - /* Optionally shift out of the way of Viridian architectural leaves. */
>> - uint32_t base = is_viridian_domain(d) ? 0x40000100 : 0x40000000;
>> + /*
>> + * Optionally shift out of the way of Viridian or VMware
>> + * architectural leaves.
>> + */
>> + uint32_t base = is_viridian_domain(d) | is_vmware_domain(d) ?
>> + 0x40000100 : 0x40000000;
> Again, brackets please for binary operators. (We have had one recent
> slipup because of the precedence of | and ?:)
I do not think they are needed, but can add them. A separate
thread also talks about this.
> Furthermore, I think you mean (is_viridian_domain(d) ||
> is_vmware_domain(d)) ? 0x40000100 : 0x40000000; which allows for short
> circuiting of is_vmware_domain().
Boris Ostrovsky siad the same. I will switch to "||".
>> uint32_t limit, dummy;
>>
>> idx -= base;
>> diff --git a/xen/include/asm-x86/hvm/hvm.h b/xen/include/asm-x86/hvm/hvm.h
>> index 1123857..546210a 100644
>> --- a/xen/include/asm-x86/hvm/hvm.h
>> +++ b/xen/include/asm-x86/hvm/hvm.h
>> @@ -347,6 +347,9 @@ static inline unsigned long hvm_get_shadow_gs_base(struct vcpu *v)
>> #define is_viridian_domain(_d) \
>> (is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VIRIDIAN]))
>>
>> +#define is_vmware_domain(_d) \
>> + (is_hvm_domain(_d) && ((_d)->arch.hvm_domain.params[HVM_PARAM_VMWARE_HW]))
>> +
>> void hvm_hypervisor_cpuid_leaf(uint32_t sub_idx,
>> uint32_t *eax, uint32_t *ebx,
>> uint32_t *ecx, uint32_t *edx);
>> diff --git a/xen/include/asm-x86/hvm/vmware.h b/xen/include/asm-x86/hvm/vmware.h
>> new file mode 100644
>> index 0000000..f254106
>> --- /dev/null
>> +++ b/xen/include/asm-x86/hvm/vmware.h
>> @@ -0,0 +1,31 @@
>> +/*
>> + * asm-x86/hvm/vmware.h
>> + *
>> + * Copyright (C) 2012 Verizon Corporation
>> + *
>> + * This file is free software; you can redistribute it and/or modify it
>> + * under the terms of the GNU General Public License Version 2 (GPLv2)
>> + * as published by the Free Software Foundation.
>> + *
>> + * This file is distributed in the hope that it will be useful,
>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
>> + * General Public License for more details. <http://www.gnu.org/licenses/>.
>> + */
>> +
>> +#ifndef ASM_X86_HVM_VMWARE_H__
>> +#define ASM_X86_HVM_VMWARE_H__
> #include <xen/types.h> (IIRC) please, or uint32_t could cause a
> compilation failure given specific orderings of #include's in
> translation units.
I am fine with adding this.
-Don Slutz
> ~Andrew
>
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 06/16] xen: Convert vmware_port to xentrace usage
2014-09-12 13:10 ` Boris Ostrovsky
@ 2014-09-12 23:57 ` Don Slutz
0 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-12 23:57 UTC (permalink / raw)
To: Boris Ostrovsky, Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/12/14 09:10, Boris Ostrovsky wrote:
> On 09/11/2014 02:36 PM, Don Slutz wrote:
>> @@ -2080,6 +2080,14 @@ static void svm_vmexit_gp_intercept(struct
>> cpu_user_regs *regs,
>> inst_len = __get_instruction_length_from_list(
>> v, list, ARRAY_SIZE(list), 0);
>> + if ( hvm_long_mode_enabled(v) )
>> + HVMTRACE_LONG2_C4D(TRAP_GP, inst_len, regs->rdx,
>> + TRC_PAR_LONG(vmcb->exitinfo1),
>> + TRC_PAR_LONG(vmcb->exitinfo2));
>> + else
>> + HVMTRACE_C4D(TRAP_GP, inst_len, vmcb->exitinfo1,
>> vmcb->exitinfo1,
>> + vmcb->exitinfo2);
>
> exitinfo1 is used twice. Did you mean regs->rdx for one of them?
>
Yes. Will fix.
Thanks,
-Don Slutz
> -boris
>
>> +
>> rc = vmport_gp_check(regs, v, inst_len, inst_addr,
>> vmcb->exitinfo1, vmcb->exitinfo2);
>> if ( !rc )
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 07/16] tools: Convert vmware_port to xentrace usage
2014-09-12 13:15 ` Boris Ostrovsky
@ 2014-09-13 0:01 ` Don Slutz
0 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-13 0:01 UTC (permalink / raw)
To: Boris Ostrovsky, Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/12/14 09:15, Boris Ostrovsky wrote:
> On 09/11/2014 02:36 PM, Don Slutz wrote:
>> Also added missing TRAP_DEBUG & VLAPIC.
>>
>> Signed-off-by: Don Slutz <dslutz@verizon.com>
>> ---
>> tools/xentrace/formats | 13 +++++++++++++
>> 1 file changed, 13 insertions(+)
>>
>> diff --git a/tools/xentrace/formats b/tools/xentrace/formats
>> index da658bf..b403c54 100644
>> --- a/tools/xentrace/formats
>> +++ b/tools/xentrace/formats
>> @@ -79,6 +79,19 @@
>> 0x00082020 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) INTR_WINDOW [ value
>> = 0x%(1)08x ]
>> 0x00082021 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) NPF [ gpa =
>> 0x%(2)08x%(1)08x mfn = 0x%(4)08x%(3)08x qual = 0x%(5)04x p2mt =
>> 0x%(6)04x ]
>> 0x00082023 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP [
>> vector = 0x%(1)02x ]
>> +0x00082024 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_DEBUG [
>> exit_qualification = 0x%(1)08x ]
>> +0x00082025 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VLAPIC
>> +0x00082026 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP [
>> inst_len = %(1)d edx = 0x%(2)08x exitinfo1 = 0x%(3)08x exitinfo2 =
>> 0x%(4)08x ]
>> +0x00082126 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP [
>> inst_len = %(1)d edx = 0x%(2)08x exitinfo1 = 0x%(4)08x%(3)08x
>> exitinfo2 = 0x%(6)08x%(5)08x ]
>> +0x00082027 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP_UNKNOWN [ rc
>> = %(1)d eax = 0x%(2)08x ebx = 0x%(3)08x ecx = 0x%(4)08x inst_addr =
>> 0x%(5)08x ]
>> +0x00082127 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) TRAP_GP_UNKNOWN [ rc
>> = %(1)d eax = 0x%(2)08x ebx = 0x%(3)08x ecx = 0x%(4)08x inst_addr =
>> 0x%(6)08x%(5)08x ]
>> +0x00082028 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_READ_BEFORE [
>> eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi =
>> 0x%(5)08x edi = 0x%(6)08x ]
>> +0x00082228 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_WRITE_BEFORE [
>> eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi =
>> 0x%(5)08x edi = 0x%(6)08x ]
>> +0x00082029 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_READ_AFTER [
>> eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi =
>> 0x%(5)08x edi = 0x%(6)08x ]
>> +0x00082229 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_WRITE_AFTER [
>> eax = 0x%(1)08x ebx = 0x%(2)08x ecx = 0x%(3)08x edx = 0x%(4)08x esi =
>> 0x%(5)08x edi = 0x%(6)08x ]
>> +0x0008202a CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_BAD [ dir =
>> %(1)d bytes = 0x%(2)d eax = 0x%(3)08x eip = 0x%(4)08x ]
>
> 'bytes = 0x%(2)d' or 'bytes = %(2)d' ?
>
It should match the 64 bit one. 'bytes = %(2)d'. Will fix.
Thanks,
-Don Slutz
> -boris
>
>> +0x0008212a CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_BAD [ dir =
>> %(1)d bytes = %(2)d eax = 0x%(3)08x rip = 0x%(5)08x%(4)08x ]
>> +0x0008202b CPU%(cpu)d %(tsc)d (+%(reltsc)8d) VMPORT_UNKNOWN [
>> bytes << 8 + dir = 0x%(1)03x cmd = 0x%(2)x cmd = %(2)d ebx =
>> 0x%(3)08x ecx = 0x%(4)08x esi = 0x%(5)08x edi = 0x%(6)08x ]
>> 0x0010f001 CPU%(cpu)d %(tsc)d (+%(reltsc)8d)
>> page_grant_map [ domid = %(1)d ]
>> 0x0010f002 CPU%(cpu)d %(tsc)d (+%(reltsc)8d) page_grant_unmap
>> [ domid = %(1)d ]
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-12 17:46 ` Don Slutz
@ 2014-09-15 7:42 ` Jan Beulich
2014-09-17 15:41 ` Don Slutz
0 siblings, 1 reply; 51+ messages in thread
From: Jan Beulich @ 2014-09-15 7:42 UTC (permalink / raw)
To: Don Slutz
Cc: Jun Nakajima, Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
xen-devel, Eddie Dong, Aravind Gopalakrishnan,
Suravee Suthikulpanit, Boris Ostrovsky
>>> On 12.09.14 at 19:46, <dslutz@verizon.com> wrote:
> On 09/12/14 05:49, Jan Beulich wrote:
>>>>> On 11.09.14 at 21:49, <andrew.cooper3@citrix.com> wrote:
>>>> + case 0x10:
>>>> + /* (Virtual) TSC frequency in kHz. */
>>>> + *eax = d->arch.tsc_khz;
>>>> + /* (Virtual) Bus (local apic timer) frequency in kHz. */
>>>> + *ebx = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
>>> At least 1 pair of brackets please, especially as the placement of
>>> brackets affects the result of this particular calculation.
>> Or simply eliminate one of the divisions using
>> "1000000ull / APIC_BUS_CYCLE_NS".
>
> I am totally confused. I am happy to go with Jan's version.
>
> The confusion is that I get the same answer all the ways I try.
Hmm - this ...
> ebx1 = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
> ebx2 = (1000000000ull / APIC_BUS_CYCLE_NS) / 1000ull;
> ebx3 = 1000000000ull / (APIC_BUS_CYCLE_NS * 1000ull);
... clearly indicates the contrary: You converted to mutiplication
here, when the respective possibility of putting parentheses here
would have been
ebx3 = 1000000000ull / (APIC_BUS_CYCLE_NS / 1000ull);
which I'm sure you agree won't produce the same result. But yes,
the language implies parentheses this way
ebx2 = (1000000000ull / APIC_BUS_CYCLE_NS) / 1000ull;
so the original expression without them was correct, just
slightly ambiguous.
Jan
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-12 13:08 ` Boris Ostrovsky
@ 2014-09-16 12:08 ` Slutz, Donald Christopher
2014-09-17 15:56 ` Boris Ostrovsky
0 siblings, 1 reply; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-16 12:08 UTC (permalink / raw)
To: Boris Ostrovsky
Cc: Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, Jun Nakajima, Eddie Dong, Ian Jackson,
xen-devel, George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Suravee Suthikulpanit
On 09/12/14 09:08, Boris Ostrovsky wrote:
> On 09/11/2014 02:36 PM, Don Slutz wrote:
>> int __get_instruction_length_from_list(struct vcpu *v,
>> - const enum instruction_index *list, unsigned int list_count)
>> + const enum instruction_index
>> *list,
>> + unsigned int list_count,
>> + bool_t err_rpt)
>> {
>> struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>> unsigned int i, j, inst_len = 0;
>> @@ -211,10 +222,13 @@ int __get_instruction_length_from_list(struct
>> vcpu *v,
>> mismatch: ;
>> }
>> - gdprintk(XENLOG_WARNING,
>> - "%s: Mismatch between expected and actual instruction
>> bytes: "
>> - "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>> - hvm_inject_hw_exception(TRAP_gp_fault, 0);
>> + if ( err_rpt )
>> + {
>> + gdprintk(XENLOG_WARNING,
>> + "%s: Mismatch between expected and actual
>> instruction bytes: "
>> + "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>> + hvm_inject_hw_exception(TRAP_gp_fault, 0);
>> + }
>> return 0;
>> done:
>> diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
>> index b5188e6..9e14d2a 100644
>> --- a/xen/arch/x86/hvm/svm/svm.c
>> +++ b/xen/arch/x86/hvm/svm/svm.c
>> @@ -59,6 +59,7 @@
>> #include <public/sched.h>
>> #include <asm/hvm/vpt.h>
>> #include <asm/hvm/trace.h>
>> +#include <asm/hvm/vmport.h>
>> #include <asm/hap.h>
>> #include <asm/apic.h>
>> #include <asm/debugger.h>
>> @@ -2065,6 +2066,38 @@ svm_vmexit_do_vmsave(struct vmcb_struct *vmcb,
>> return;
>> }
>> +static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
>> + struct vcpu *v)
>> +{
>> + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>> + unsigned long inst_len;
>> + unsigned long inst_addr = svm_rip2pointer(v);
>> + int rc;
>> + static const enum instruction_index list[] = {
>> + INSTR_INL_DX, INSTR_INB_DX, INSTR_OUTL_DX, INSTR_OUTB_DX
>> + };
>> +
>> + inst_len = __get_instruction_length_from_list(
>> + v, list, ARRAY_SIZE(list), 0);
>
> I should have asked earlier but I don't think I understand why the
> last argument here is 0 (and therefore why you have this last argument
> at all).
>
> Because whether or not you are warning in
> __get_instruction_length_from_list() it will still return 0. And that,
> in turn, will cause vmport_gp_check() to return an error. And then you
> will print another warning in VMPORT_LOG. So there is a warning anyway.
>
A key part that you appear to have missed is that
__get_instruction_length_from_list() uses gdprintk(XENLOG_WARNING,...
but VMPORT_DBG_LOG is only available in debug=y builds. So the new
last argument is used to control this. Since this change includes
enabling #GP vmexits, it is now possible for ring 3 users to generate a
large volume of these which with gdprintk() can flood the console.
> Second, since this handler appears to be handling #GP only for VMware
> guest we should make sure that it is not executed for any other guest.
> You do now condition intercept got #GP for such guests only but I
> still think having a check here is worth doing. Maybe a BUG() or
> ASSERT()?
>
> The same comments are applicable to VMX code, I suspect.
>
I will change the check in vmport_gp_check on is_vmware_port_enabled
into an ASSERT() so both SVM and VMX will be covered.
>> +
>> + rc = vmport_gp_check(regs, v, inst_len, inst_addr,
>> + vmcb->exitinfo1, vmcb->exitinfo2);
>> + if ( !rc )
>> + __update_guest_eip(regs, inst_len);
>> + else
>> + {
>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_UNKNOWN,
>> + "gp: rc=%d ei1=0x%lx ei2=0x%lx ip=%"PRIx64
>> + " (0x%lx,%ld) ax=%"PRIx64" bx=%"PRIx64"
>> cx=%"PRIx64
>> + " dx=%"PRIx64" si=%"PRIx64" di=%"PRIx64, rc,
>> + (unsigned long)vmcb->exitinfo1,
>> + (unsigned long)vmcb->exitinfo2, regs->rip,
>> inst_addr,
>> + inst_len, regs->rax, regs->rbx, regs->rcx,
>> regs->rdx,
>> + regs->rsi, regs->rdi);
>> + hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
>> + }
>> +}
>> +
>
> ...
>
>> +
>> +int vmport_gp_check(struct cpu_user_regs *regs, struct vcpu *v,
>> + unsigned long inst_len, unsigned long inst_addr,
>> + unsigned long ei1, unsigned long ei2)
>> +{
>> + if ( !v->domain->arch.hvm_domain.is_vmware_port_enabled )
>> + return 10;
I.E. ASSERT(v->domain->arch.hvm_domain.is_vmware_port_enabled);
>> +
>> + if ( inst_len && inst_len <= 2 && get_low_bits(regs->rdx) ==
>> BDOOR_PORT &&
>> + ei1 == 0 && ei2 == 0 && regs->error_code == 0 &&
>> + (uint32_t)regs->rax == BDOOR_MAGIC )
>> + {
>> + int i = 0;
>> + uint32_t val;
>> + uint32_t byte_cnt = 4;
>> + unsigned char bytes[2];
>> + unsigned int fetch_len;
>> + int frc;
>> + int rc;
>> +
>> + /*
>> + * Fetch up to the next page break; we'll fetch from the
>> + * next page later if we have to.
>> + */
>> + fetch_len = min_t(unsigned int, inst_len,
>> + PAGE_SIZE - (inst_addr & ~PAGE_MASK));
>> + frc = hvm_fetch_from_guest_virt_nofault(bytes, inst_addr,
>> fetch_len,
>> + PFEC_page_present);
>> + if ( frc != HVMCOPY_okay )
>> + {
>> + gdprintk(XENLOG_WARNING,
>> + "Bad instruction fetch at %#lx (frc=%d il=%lu
>> fl=%u)\n",
>> + (unsigned long) inst_addr, frc, inst_len,
>> fetch_len);
>> + return 11;
>> + }
>> + if ( bytes[0] == 0x66 ) /* operand size prefix */
>> + {
>> + byte_cnt = 2;
>> + i = 1;
>> + if ( fetch_len != inst_len )
>> + {
>> + frc = hvm_fetch_from_guest_virt_nofault(&bytes[1],
>> + inst_addr + 1, 1,
>> + PFEC_page_present);
>> + if ( frc != HVMCOPY_okay )
>> + {
>> + gdprintk(XENLOG_WARNING,
>> + "Bad instruction fetch at %#lx + 1
>> (frc=%d)\n",
>> + (unsigned long) inst_addr, frc);
>> + return 12;
>> + }
>> + }
>> + }
>> + if ( bytes[i] == 0xed ) /* in (%dx),%eax or in (%dx),%ax */
>> + {
>> + rc = vmport_ioport(IOREQ_READ, BDOOR_PORT, byte_cnt, &val);
>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
>> + "gp: VMwareIn rc=%d ip=%"PRIx64"
>> byte_cnt=%d ax=%"
>> + PRIx64" bx=%"PRIx64" cx=%"PRIx64"
>> dx=%"PRIx64
>> + " si=%"PRIx64" di=%"PRIx64, rc,
>> + inst_addr, byte_cnt, regs->rax, regs->rbx,
>> + regs->rcx, regs->rdx, regs->rsi, regs->rdi);
>> + return rc;
>> + }
>> + else if ( bytes[i] == 0xec ) /* in (%dx),%al */
>> + {
>> + rc = vmport_ioport(IOREQ_READ, BDOOR_PORT, 1, &val);
>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
>> + "gp: VMwareIn rc=%d ip=%"PRIx64"
>> byte_cnt=1 ax=%"
>> + PRIx64" bx=%"PRIx64" cx=%"PRIx64"
>> dx=%"PRIx64
>> + " si=%"PRIx64" di=%"PRIx64, rc,
>> + inst_addr, regs->rax, regs->rbx, regs->rcx,
>> + regs->rdx, regs->rsi, regs->rdi);
>> + return rc;
>> + }
>> + else if ( bytes[i] == 0xef ) /* out %eax,(%dx) or out
>> %ax,(%dx) */
>> + {
>> + rc = vmport_ioport(IOREQ_WRITE, BDOOR_PORT, byte_cnt,
>> &val);
>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
>> + "gp: VMwareOut rc=%d ip=%"PRIx64"
>> byte_cnt=%d ax=%"
>> + PRIx64" bx=%"PRIx64" cx=%"PRIx64"
>> dx=%"PRIx64
>> + " si=%"PRIx64" di=%"PRIx64, rc,
>> + inst_addr, byte_cnt, regs->rax, regs->rbx,
>> + regs->rcx, regs->rdx, regs->rsi, regs->rdi);
>> + return rc;
>> + }
>> + else if ( bytes[i] == 0xee ) /* out %al,(%dx) */
>> + {
>> + rc = vmport_ioport(IOREQ_WRITE, BDOOR_PORT, 1, &val);
>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_VMWARE_AFTER,
>> + "gp: VMwareOut rc=%d ip=%"PRIx64"
>> byte_cnt=1 ax=%"
>> + PRIx64" bx=%"PRIx64" cx=%"PRIx64"
>> dx=%"PRIx64
>> + " si=%"PRIx64" di=%"PRIx64, rc,
>> + inst_addr, regs->rax, regs->rbx, regs->rcx,
>> + regs->rdx, regs->rsi, regs->rdi);
>> + return rc;
>> + }
>> + else
>> + {
>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_FAIL_RD_INST,
>> + "gp: VMware? lip=%"PRIx64"[%d]=>0x%x(%ld)
>> ax=%"
>> + PRIx64" bx=%"PRIx64" cx=%"PRIx64"
>> dx=%"PRIx64
>> + " si=%"PRIx64" di=%"PRIx64,
>> + inst_addr, i, bytes[i], inst_len, regs->rax,
>> + regs->rbx, regs->rcx, regs->rdx, regs->rsi,
>> + regs->rdi);
>> + return 13;
>> + }
>> + }
>> + return 14;
>
> The return values should be defined as macros --- otherwise they look
> like some magic integers.
>
Ok, will add a set of #define for them.
-Don Slutz
> -boris
>
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc
2014-09-12 14:27 ` Jan Beulich
@ 2014-09-16 12:38 ` Slutz, Donald Christopher
2014-09-16 12:46 ` Jan Beulich
0 siblings, 1 reply; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-16 12:38 UTC (permalink / raw)
To: Jan Beulich
Cc: Jun Nakajima, Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
xen-devel, Eddie Dong, Aravind Gopalakrishnan,
Suravee Suthikulpanit, Boris Ostrovsky
On 09/12/14 10:27, Jan Beulich wrote:
>>>> On 12.09.14 at 15:37, <boris.ostrovsky@oracle.com> wrote:
>> On 09/11/2014 02:36 PM, Don Slutz wrote:
>>> +static inline uint16_t get_low_bits(uint32_t bits)
>>> +{
>>> + return bits & 0xffff;
>>> +}
>>> +
>>> +static inline uint16_t get_high_bits(uint32_t bits)
>>> +{
>>> + return bits >> 16;
>>> +}
>>> +
>>> +static inline uint32_t set_high_bits(uint32_t b, uint32_t val)
>>> +{
>>> + return (val << 16) | get_low_bits(b);
>>> +}
> The names of all three functions (at least one of which I think I saw
> in a reply to an earlier patch - code duplication?) are bogus: How
> many high or low bits (and for the "high" case out of how many) are
> we talking about here?
I have no issue with changing the names. How does
s/get_low_bits/get_low_16bits_of_32/
s/get_high_bits/get_high_16bits_of_32/
s/set_high_bits/set_high_16bits_of_32/
look as a fix?
-Don Slutz
>>> +static inline void set_status(struct cpu_user_regs *ur, uint16_t val)
>>> +{
>>> + /* VMware defines this to be only 32 bits */
>>> + ur->rcx = (val << 16) | (ur->rcx & 0xffff);
>> Are you shifting val out of its size? Or will compiler automatically
>> promote this to uint32_t? (It does for me but I don't know if this is
>> something we can rely on).
> All operands more narrow than int get promoted to int.
>
> Jan
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc
2014-09-16 12:38 ` Slutz, Donald Christopher
@ 2014-09-16 12:46 ` Jan Beulich
2014-09-16 13:47 ` Slutz, Donald Christopher
0 siblings, 1 reply; 51+ messages in thread
From: Jan Beulich @ 2014-09-16 12:46 UTC (permalink / raw)
To: Donald Christopher Slutz
Cc: Jun Nakajima, Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
xen-devel, Eddie Dong, Aravind Gopalakrishnan,
Suravee Suthikulpanit, Boris Ostrovsky
>>> On 16.09.14 at 14:38, <dslutz@verizon.com> wrote:
> On 09/12/14 10:27, Jan Beulich wrote:
>>>>> On 12.09.14 at 15:37, <boris.ostrovsky@oracle.com> wrote:
>>> On 09/11/2014 02:36 PM, Don Slutz wrote:
>>>> +static inline uint16_t get_low_bits(uint32_t bits)
>>>> +{
>>>> + return bits & 0xffff;
>>>> +}
>>>> +
>>>> +static inline uint16_t get_high_bits(uint32_t bits)
>>>> +{
>>>> + return bits >> 16;
>>>> +}
>>>> +
>>>> +static inline uint32_t set_high_bits(uint32_t b, uint32_t val)
>>>> +{
>>>> + return (val << 16) | get_low_bits(b);
>>>> +}
>> The names of all three functions (at least one of which I think I saw
>> in a reply to an earlier patch - code duplication?) are bogus: How
>> many high or low bits (and for the "high" case out of how many) are
>> we talking about here?
>
> I have no issue with changing the names. How does
>
> s/get_low_bits/get_low_16bits_of_32/
> s/get_high_bits/get_high_16bits_of_32/
> s/set_high_bits/set_high_16bits_of_32/
>
> look as a fix?
Ugly. What's wrong with dropping these wrappers? Would the
resulting code be so much worse to read?
Jan
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc
2014-09-12 13:37 ` Boris Ostrovsky
2014-09-12 14:27 ` Jan Beulich
@ 2014-09-16 13:17 ` Slutz, Donald Christopher
1 sibling, 0 replies; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-16 13:17 UTC (permalink / raw)
To: Boris Ostrovsky
Cc: Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, Jun Nakajima, Eddie Dong, Ian Jackson,
xen-devel, George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Suravee Suthikulpanit
On 09/12/14 09:37, Boris Ostrovsky wrote:
> On 09/11/2014 02:36 PM, Don Slutz wrote:
>> @@ -6142,6 +6148,43 @@ long do_hvm_op(unsigned long op,
>> XEN_GUEST_HANDLE_PARAM(void) arg)
>> break;
>> }
>> + case HVMOP_get_vmport_guest_info:
>> + case HVMOP_set_vmport_guest_info:
>> + {
>> + struct xen_hvm_vmport_guest_info a;
>> + struct domain *d;
>> +
>> + if ( copy_from_guest(&a, arg, 1) )
>> + return -EFAULT;
>> +
>> + rc = vmport_rpc_hvmop_precheck(op, &a);
>> + if ( rc )
>> + return rc;
>> +
>> + d = rcu_lock_domain_by_any_id(a.domid);
>> + if ( d == NULL )
>> + return rc;
>> +
>> + rc = -EINVAL;
>> + if ( !is_hvm_domain(d) )
>> + goto param_fail9;
>
> Do we need this check? If you are concerned about PVH then checking
> explicitly for that may be better (and probably do it first thing in
> the case clause)
>
I am not sure. This is copied from HVMOP_inject_trap; and a lot of the
other cases have the same test. I do not see that PVH needs to be
excluded. It is the 1st thing after d has a value.
>> +
>> + rc = xsm_hvm_param(XSM_TARGET, d, op);
>> + if ( rc )
>> + goto param_fail9;
>> +
>> + rc = vmport_rpc_hvmop_do(d, op, &a);
>> + if ( rc )
>> + goto param_fail9;
>> +
>> + if ( op == HVMOP_get_vmport_guest_info )
>> + rc = copy_to_guest(arg, &a, 1) ? -EFAULT : 0;
>> +
>> + param_fail9:
>> + rcu_unlock_domain(d);
>> + break;
>> + }
>> +
>> default:
>> {
>> gdprintk(XENLOG_DEBUG, "Bad HVM op %ld.\n", op);
>> diff --git a/xen/arch/x86/hvm/vmware/Makefile
>> b/xen/arch/x86/hvm/vmware/Makefile
>> index cd8815b..4a14124 100644
>> --- a/xen/arch/x86/hvm/vmware/Makefile
>> +++ b/xen/arch/x86/hvm/vmware/Makefile
>> @@ -1,2 +1,3 @@
>> obj-y += cpuid.o
>> obj-y += vmport.o
>> +obj-y += vmport_rpc.o
>> diff --git a/xen/arch/x86/hvm/vmware/vmport.c
>> b/xen/arch/x86/hvm/vmware/vmport.c
>> index 26aeb37..9e308a0 100644
>> --- a/xen/arch/x86/hvm/vmware/vmport.c
>> +++ b/xen/arch/x86/hvm/vmware/vmport.c
>> @@ -139,6 +139,13 @@ int vmport_ioport(int dir, uint32_t port,
>> uint32_t bytes, uint32_t *val)
>> /* maxTimeLag */
>> regs->rcx = 0;
>> break;
>> + case BDOOR_CMD_MESSAGE:
>> + if ( !is_pv_vcpu(current) )
>> + {
>> + /* Only supported for non pv domains */
>
> PV vs. HVM vs. PVH. So probably 'if(is_hvm_vcpu)'?
I see no reason to exclude PVH. Will change to has_hvm_container_vcpu()
-Don Slutz
(rest on different thread.)
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc
2014-09-16 12:46 ` Jan Beulich
@ 2014-09-16 13:47 ` Slutz, Donald Christopher
0 siblings, 0 replies; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-16 13:47 UTC (permalink / raw)
To: Jan Beulich
Cc: Jun Nakajima, Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
xen-devel, Eddie Dong, Aravind Gopalakrishnan,
Suravee Suthikulpanit, Boris Ostrovsky
On 09/16/14 08:46, Jan Beulich wrote:
>>>> On 16.09.14 at 14:38, <dslutz@verizon.com> wrote:
>> On 09/12/14 10:27, Jan Beulich wrote:
>>>>>> On 12.09.14 at 15:37, <boris.ostrovsky@oracle.com> wrote:
>>>> On 09/11/2014 02:36 PM, Don Slutz wrote:
>>>>> +static inline uint16_t get_low_bits(uint32_t bits)
>>>>> +{
>>>>> + return bits & 0xffff;
>>>>> +}
>>>>> +
>>>>> +static inline uint16_t get_high_bits(uint32_t bits)
>>>>> +{
>>>>> + return bits >> 16;
>>>>> +}
>>>>> +
>>>>> +static inline uint32_t set_high_bits(uint32_t b, uint32_t val)
>>>>> +{
>>>>> + return (val << 16) | get_low_bits(b);
>>>>> +}
>>> The names of all three functions (at least one of which I think I saw
>>> in a reply to an earlier patch - code duplication?) are bogus: How
>>> many high or low bits (and for the "high" case out of how many) are
>>> we talking about here?
>> I have no issue with changing the names. How does
>>
>> s/get_low_bits/get_low_16bits_of_32/
>> s/get_high_bits/get_high_16bits_of_32/
>> s/set_high_bits/set_high_16bits_of_32/
>>
>> look as a fix?
> Ugly. What's wrong with dropping these wrappers? Would the
> resulting code be so much worse to read?
I guess not. I can go that way.
-Don Slutz
> Jan
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 11/16] Add live migration of VMware's hyper-call RPC
2014-09-12 13:54 ` Boris Ostrovsky
@ 2014-09-16 15:48 ` Don Slutz
0 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-16 15:48 UTC (permalink / raw)
To: Boris Ostrovsky, Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
Jun Nakajima, Eddie Dong, Ian Jackson, Tim Deegan, George Dunlap,
Aravind Gopalakrishnan, Jan Beulich, Andrew Cooper,
Suravee Suthikulpanit
On 09/12/14 09:54, Boris Ostrovsky wrote:
> On 09/11/2014 02:36 PM, Don Slutz wrote:
>> The VMware's hyper-call state is included in live migration and
>> save/restore. Because the max size of the VMware guestinfo is
>> large, then data is compressed and expanded in the
>> vmport_save_domain_ctxt and vmport_load_domain_ctxt.
>>
>> Signed-off-by: Don Slutz <dslutz@verizon.com>
>> ---
>>
>> +
>> + for ( i = 0; i < ctxt->used_guestinfo; i++ )
>> + {
>> + vmport_guestinfo_t *vg = vs->guestinfo[i];
>> +
>> + if ( vg && vg->key_len )
>> + {
>> + guestinfo_size += sizeof(vg->key_len) +
>> sizeof(vg->val_len) +
>> + vg->key_len + vg->val_len;
>> + used_guestinfo++;
>> + ASSERT(sizeof(vg->key_len) == 1);
>> + *p++ = (char) vg->key_len;
>> + ASSERT(sizeof(vg->val_len) == 1);
>> + *p++ = (char) vg->val_len;
>> + if ( vg->key_len )
>
> You ASSERTed that vg->key_len is 1 so you may not need the 'if'. (And
> in NDEBUG version the worst that could happen is you do memcpy(,,0),
> which is safe).
>
You seem to have missed the sizeof() in there. So vg->key_len can be
zero. But I will agree
that the if could be dropped.
>> + {
>> + memcpy(p, vg->key_data, vg->key_len);
>> + p += vg->key_len;
>> + if ( vg->val_len )
>
> Same here.
>
Yes, the if is not needed. Can drop if requested.
>> + {
>> + memcpy(p, vg->val_data, vg->val_len);
>> + p += vg->val_len;
>> + }
>> + }
>> + }
>> + }
>> + ctxt->used_guestinfo = used_guestinfo;
>> +
>> + for ( i = 0; i < ctxt->used_guestinfo_jumbo; i++ )
>> + {
>> + vmport_guestinfo_jumbo_t *vgj =
>> + vs->guestinfo_jumbo[i];
>> + if ( vgj && vgj->key_len )
>> + {
>> + guestinfo_size += sizeof(vgj->key_len) +
>> sizeof(vgj->val_len) +
>> + vgj->key_len + vgj->val_len;
>> + used_guestinfo_jumbo++;
>> + ASSERT(sizeof(vgj->key_len) == 1);
>> + *p++ = (char) vgj->key_len;
>> + memcpy(p, &vgj->val_len, sizeof(vgj->val_len));
>> + p += sizeof(vgj->val_len);
>> + if ( vgj->key_len )
>
> And here.
>
> And do you need ASSERT(vgj->val_len ==1) as well? Also, In
> vmport_load_domain_ctxt() you don't seem to be making this assertions.
> I don't know whether this is on purpose.
>
In this case sizeof(vgj->val_len) != 1.
I will add ASSERTs there also.
>> +
>> + if ( !vs )
>> + return -ENOMEM;
>> +
>> + /* Customized checking for entry since our entry is of variable
>> length */
>> + desc = (struct hvm_save_descriptor *)&h->data[h->cur];
>> + if ( sizeof(*desc) > h->size - h->cur )
>> + {
>> + printk(XENLOG_G_WARNING
>> + "HVM%d restore: not enough data left to read descriptor"
>> + "for type %lu\n", d->domain_id,
>> + HVM_SAVE_CODE(VMPORT));
>> + return -1;
>
> Since in some cases you are returning proper error codes you should
> return them everywhere.
>
Will adjust to proper error codes.
-Don Slutz
> -boris
>
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 02/16] tools: Add vmware_hw support
2014-09-11 21:09 ` Andrew Cooper
@ 2014-09-16 16:20 ` Don Slutz
0 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-16 16:20 UTC (permalink / raw)
To: Andrew Cooper, Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
George Dunlap, Ian Jackson, Eddie Dong, Tim Deegan, Jan Beulich,
Aravind Gopalakrishnan, Jun Nakajima, Boris Ostrovsky,
Suravee Suthikulpanit
On 09/11/14 17:09, Andrew Cooper wrote:
> On 11/09/2014 19:36, Don Slutz wrote:
>> This is used to set HVM_PARAM_VMWARE_HW. It is set to the VMware
>> virtual hardware version.
>>
>> Currently 0, 3-4, 6-11 are good values. However the code only
>> checks for == 0 or != 0.
>>
>> If non-zero then
>> default VGA to VMware's VGA.
>>
>> Also now allows vga=vmware
>>
>> Signed-off-by: Don Slutz <dslutz@verizon.com>
>> ---
>> docs/man/xl.cfg.pod.5 | 21 +++++++++++++++++++--
>> docs/misc/hypervisor-cpuid.markdown | 29 +++++++++++++++++++++++++++++
>> tools/libxc/xc_domain_restore.c | 14 ++++++++++++++
>> tools/libxc/xc_domain_save.c | 11 +++++++++++
>> tools/libxc/xg_save_restore.h | 2 ++
>> tools/libxl/libxl.h | 10 ++++++++++
>> tools/libxl/libxl_create.c | 4 +++-
>> tools/libxl/libxl_dm.c | 10 +++++++++-
>> tools/libxl/libxl_dom.c | 2 ++
>> tools/libxl/libxl_types.idl | 2 ++
>> tools/libxl/xl_cmdimpl.c | 11 ++++++++++-
>> 11 files changed, 111 insertions(+), 5 deletions(-)
>> create mode 100644 docs/misc/hypervisor-cpuid.markdown
>>
>> diff --git a/docs/man/xl.cfg.pod.5 b/docs/man/xl.cfg.pod.5
>> index 517ae2f..367b401 100644
>> --- a/docs/man/xl.cfg.pod.5
>> +++ b/docs/man/xl.cfg.pod.5
>> @@ -1147,6 +1147,23 @@ some other Operating Systems and in some circumstance can prevent
>> Xen's own paravirtualisation interfaces for HVM guests from being
>> used.
>>
>> +=item B<vmware_hw=NUMBER>
>> +
>> +Turns on or off the exposure of VMware cpuid. The number is the
>> +VMware's hardware version number, where 0 is off. If not zero it
>> +changes the default VGA to VMware's VGA.
>> +
>> +The hardware version number (vmware_hw) come from VMware config files.
>> +
>> +=over 4
>> +
>> +In a .vmx it is virtualHW.version
>> +
>> +In a .ovf it is part of the value of vssd:VirtualSystemType.
>> +For vssd:VirtualSystemType == vmx-07, vmware_hw = 7.
>> +
>> +=back
>> +
>> =back
>>
>> =head3 Emulated VGA Graphics Device
>> @@ -1185,8 +1202,8 @@ This option is deprecated, use vga="stdvga" instead.
>>
>> =item B<vga="STRING">
>>
>> -Selects the emulated video card (none|stdvga|cirrus).
>> -The default is cirrus.
>> +Selects the emulated video card (none|stdvga|cirrus|vmware).
>> +The default is cirrus (or vmware if B<vmware_hw> is not zero).
>>
>> =item B<vnc=BOOLEAN>
>>
>> diff --git a/docs/misc/hypervisor-cpuid.markdown b/docs/misc/hypervisor-cpuid.markdown
>> new file mode 100644
>> index 0000000..4199d57
>> --- /dev/null
>> +++ b/docs/misc/hypervisor-cpuid.markdown
>> @@ -0,0 +1,29 @@
>> +Hypervisor Cpuid
>> +================
>> +
>> +The support of hypervisor cpuid leaves has not been agreed to.
>> +Other then the range 0x40000000 to 0x400000ff can be used by
>> +hypervisors.
>> +
>> +MicroSoft Hyper-V (AKA viridian) currently must be at 0x40000000.
>> +
>> +VMware currently must be at 0x40000000.
>> +
>> +KVM currently must be at 0x40000000 (from Seabios).
>> +
>> +Seabios will find Xen at 0x40000000, 0x40000100, 0x40000200 ..
>> +0x40010000.
> Anything looking for Xen according to the Xen cpuid instructions in the
> public header files will find Xen in this manner; hvmloader and
> HVMLinux to name but a few. I don't think Seabios wants calling out
> specifically.
>
> Perhaps "Xen can be found at the first otherwise unused 0x100 aligned
> offset between 0x40000000 and 0x40010000"?
Looks fine to me. Will change.
-Don Slutz
> ~Andrew
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-11 21:22 ` Andrew Cooper
@ 2014-09-16 16:20 ` Don Slutz
0 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-16 16:20 UTC (permalink / raw)
To: Andrew Cooper, Don Slutz, xen-devel
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
George Dunlap, Ian Jackson, Eddie Dong, Tim Deegan, Jan Beulich,
Aravind Gopalakrishnan, Jun Nakajima, Boris Ostrovsky,
Suravee Suthikulpanit
On 09/11/14 17:22, Andrew Cooper wrote:
> On 11/09/2014 19:36, Don Slutz wrote:
>> --- a/xen/arch/x86/hvm/vmx/vmx.c
>> +++ b/xen/arch/x86/hvm/vmx/vmx.c
>> @@ -44,6 +44,7 @@
>> #include <asm/hvm/support.h>
>> #include <asm/hvm/vmx/vmx.h>
>> #include <asm/hvm/vmx/vmcs.h>
>> +#include <asm/hvm/vmport.h>
>> #include <public/sched.h>
>> #include <public/hvm/ioreq.h>
>> #include <asm/hvm/vpic.h>
>> @@ -1279,6 +1280,8 @@ static void vmx_update_guest_cr(struct vcpu *v, unsigned int cr)
>> v->arch.hvm_vmx.exception_bitmap = HVM_TRAP_MASK
>> | (paging_mode_hap(v->domain) ?
>> 0 : (1U << TRAP_page_fault))
>> + | v->domain->arch.hvm_domain.is_vmware_port_enabled ?
>> + (1U << TRAP_gp_fault) : 0
> I am fairly certain that you need some brackets here.
I will add hem.
-Don Slutz
> ~Andrew
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 01/16] xen: Add support for VMware cpuid leaves
2014-09-15 7:42 ` Jan Beulich
@ 2014-09-17 15:41 ` Don Slutz
0 siblings, 0 replies; 51+ messages in thread
From: Don Slutz @ 2014-09-17 15:41 UTC (permalink / raw)
To: Jan Beulich, Don Slutz
Cc: Jun Nakajima, Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
xen-devel, Eddie Dong, Aravind Gopalakrishnan,
Suravee Suthikulpanit, Boris Ostrovsky
On 09/15/14 03:42, Jan Beulich wrote:
>>>> On 12.09.14 at 19:46, <dslutz@verizon.com> wrote:
>> On 09/12/14 05:49, Jan Beulich wrote:
>>>>>> On 11.09.14 at 21:49, <andrew.cooper3@citrix.com> wrote:
>>>>> + case 0x10:
>>>>> + /* (Virtual) TSC frequency in kHz. */
>>>>> + *eax = d->arch.tsc_khz;
>>>>> + /* (Virtual) Bus (local apic timer) frequency in kHz. */
>>>>> + *ebx = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
>>>> At least 1 pair of brackets please, especially as the placement of
>>>> brackets affects the result of this particular calculation.
>>> Or simply eliminate one of the divisions using
>>> "1000000ull / APIC_BUS_CYCLE_NS".
>> I am totally confused. I am happy to go with Jan's version.
>>
>> The confusion is that I get the same answer all the ways I try.
> Hmm - this ...
>
>> ebx1 = 1000000000ull / APIC_BUS_CYCLE_NS / 1000ull;
>> ebx2 = (1000000000ull / APIC_BUS_CYCLE_NS) / 1000ull;
>> ebx3 = 1000000000ull / (APIC_BUS_CYCLE_NS * 1000ull);
> ... clearly indicates the contrary: You converted to mutiplication
> here, when the respective possibility of putting parentheses here
> would have been
>
> ebx3 = 1000000000ull / (APIC_BUS_CYCLE_NS / 1000ull);
>
> which I'm sure you agree won't produce the same result. But yes,
> the language implies parentheses this way
Yes, I agree. Since that is not mathematical correct either. This has
the same issue
as "a - b -c" is the same as "(a - b) - c" but not "a - (b - c)".
However the "order" can be
changed. "a / b / c" and "a / c / b" is the same.
> ebx2 = (1000000000ull / APIC_BUS_CYCLE_NS) / 1000ull;
>
> so the original expression without them was correct, just
> slightly ambiguous.
Ok, I see that preventing mistakes here can only help.
-Don Slutz
> Jan
>
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-16 12:08 ` Slutz, Donald Christopher
@ 2014-09-17 15:56 ` Boris Ostrovsky
2014-09-17 18:23 ` Slutz, Donald Christopher
0 siblings, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-17 15:56 UTC (permalink / raw)
To: Slutz, Donald Christopher
Cc: Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, Jun Nakajima, Eddie Dong, Ian Jackson,
xen-devel, George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Suravee Suthikulpanit
On 09/16/2014 08:08 AM, Slutz, Donald Christopher wrote:
> On 09/12/14 09:08, Boris Ostrovsky wrote:
>> On 09/11/2014 02:36 PM, Don Slutz wrote:
>>> int __get_instruction_length_from_list(struct vcpu *v,
>>> - const enum instruction_index *list, unsigned int list_count)
>>> + const enum instruction_index
>>> *list,
>>> + unsigned int list_count,
>>> + bool_t err_rpt)
>>> {
>>> struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>> unsigned int i, j, inst_len = 0;
>>> @@ -211,10 +222,13 @@ int __get_instruction_length_from_list(struct
>>> vcpu *v,
>>> mismatch: ;
>>> }
>>> - gdprintk(XENLOG_WARNING,
>>> - "%s: Mismatch between expected and actual instruction
>>> bytes: "
>>> - "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>>> - hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>> + if ( err_rpt )
>>> + {
>>> + gdprintk(XENLOG_WARNING,
>>> + "%s: Mismatch between expected and actual
>>> instruction bytes: "
>>> + "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>>> + hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>> + }
>>> return 0;
>>> done:
>>> diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
>>> index b5188e6..9e14d2a 100644
>>> --- a/xen/arch/x86/hvm/svm/svm.c
>>> +++ b/xen/arch/x86/hvm/svm/svm.c
>>> @@ -59,6 +59,7 @@
>>> #include <public/sched.h>
>>> #include <asm/hvm/vpt.h>
>>> #include <asm/hvm/trace.h>
>>> +#include <asm/hvm/vmport.h>
>>> #include <asm/hap.h>
>>> #include <asm/apic.h>
>>> #include <asm/debugger.h>
>>> @@ -2065,6 +2066,38 @@ svm_vmexit_do_vmsave(struct vmcb_struct *vmcb,
>>> return;
>>> }
>>> +static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
>>> + struct vcpu *v)
>>> +{
>>> + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>> + unsigned long inst_len;
>>> + unsigned long inst_addr = svm_rip2pointer(v);
>>> + int rc;
>>> + static const enum instruction_index list[] = {
>>> + INSTR_INL_DX, INSTR_INB_DX, INSTR_OUTL_DX, INSTR_OUTB_DX
>>> + };
>>> +
>>> + inst_len = __get_instruction_length_from_list(
>>> + v, list, ARRAY_SIZE(list), 0);
>> I should have asked earlier but I don't think I understand why the
>> last argument here is 0 (and therefore why you have this last argument
>> at all).
>>
>> Because whether or not you are warning in
>> __get_instruction_length_from_list() it will still return 0. And that,
>> in turn, will cause vmport_gp_check() to return an error. And then you
>> will print another warning in VMPORT_LOG. So there is a warning anyway.
>>
> A key part that you appear to have missed is that
> __get_instruction_length_from_list() uses gdprintk(XENLOG_WARNING,...
> but VMPORT_DBG_LOG is only available in debug=y builds. So the new
> last argument is used to control this. Since this change includes
> enabling #GP vmexits, it is now possible for ring 3 users to generate at
> large volume of these which with gdprintk() can flood the console.
Would it be possible to decide where and whether to print the warning
inside __get_instruction_length_from_list() as opposed to passing a new
parameter? E.g. if vmware_port_enabled is set and list includes IN/OUT
and possibly something else?
-boris
>
>
>> Second, since this handler appears to be handling #GP only for VMware
>> guest we should make sure that it is not executed for any other guest.
>> You do now condition intercept got #GP for such guests only but I
>> still think having a check here is worth doing. Maybe a BUG() or
>> ASSERT()?
>>
>> The same comments are applicable to VMX code, I suspect.
>>
> I will change the check in vmport_gp_check on is_vmware_port_enabled
> into an ASSERT() so both SVM and VMX will be covered.
>
>>> +
>>> + rc = vmport_gp_check(regs, v, inst_len, inst_addr,
>>> + vmcb->exitinfo1, vmcb->exitinfo2);
>>> + if ( !rc )
>>> + __update_guest_eip(regs, inst_len);
>>> + else
>>> + {
>>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_UNKNOWN,
>>> + "gp: rc=%d ei1=0x%lx ei2=0x%lx ip=%"PRIx64
>>> + " (0x%lx,%ld) ax=%"PRIx64" bx=%"PRIx64"
>>> cx=%"PRIx64
>>> + " dx=%"PRIx64" si=%"PRIx64" di=%"PRIx64, rc,
>>> + (unsigned long)vmcb->exitinfo1,
>>> + (unsigned long)vmcb->exitinfo2, regs->rip,
>>> inst_addr,
>>> + inst_len, regs->rax, regs->rbx, regs->rcx,
>>> regs->rdx,
>>> + regs->rsi, regs->rdi);
>>> + hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
>>> + }
>>> +}
>>> +
>> .
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-17 15:56 ` Boris Ostrovsky
@ 2014-09-17 18:23 ` Slutz, Donald Christopher
2014-09-18 9:14 ` Jan Beulich
2014-09-18 22:53 ` Boris Ostrovsky
0 siblings, 2 replies; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-17 18:23 UTC (permalink / raw)
To: Boris Ostrovsky
Cc: Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, Jun Nakajima, Eddie Dong, Ian Jackson,
xen-devel, George Dunlap, Aravind Gopalakrishnan, Jan Beulich,
Andrew Cooper, Suravee Suthikulpanit
On 09/17/14 11:56, Boris Ostrovsky wrote:
>
> On 09/16/2014 08:08 AM, Slutz, Donald Christopher wrote:
>> On 09/12/14 09:08, Boris Ostrovsky wrote:
>>> On 09/11/2014 02:36 PM, Don Slutz wrote:
>>>> int __get_instruction_length_from_list(struct vcpu *v,
>>>> - const enum instruction_index *list, unsigned int list_count)
>>>> + const enum instruction_index
>>>> *list,
>>>> + unsigned int list_count,
>>>> + bool_t err_rpt)
>>>> {
>>>> struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>>> unsigned int i, j, inst_len = 0;
>>>> @@ -211,10 +222,13 @@ int __get_instruction_length_from_list(struct
>>>> vcpu *v,
>>>> mismatch: ;
>>>> }
>>>> - gdprintk(XENLOG_WARNING,
>>>> - "%s: Mismatch between expected and actual instruction
>>>> bytes: "
>>>> - "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>>>> - hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>>> + if ( err_rpt )
>>>> + {
>>>> + gdprintk(XENLOG_WARNING,
>>>> + "%s: Mismatch between expected and actual
>>>> instruction bytes: "
>>>> + "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>>>> + hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>>> + }
>>>> return 0;
>>>> done:
>>>> diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
>>>> index b5188e6..9e14d2a 100644
>>>> --- a/xen/arch/x86/hvm/svm/svm.c
>>>> +++ b/xen/arch/x86/hvm/svm/svm.c
>>>> @@ -59,6 +59,7 @@
>>>> #include <public/sched.h>
>>>> #include <asm/hvm/vpt.h>
>>>> #include <asm/hvm/trace.h>
>>>> +#include <asm/hvm/vmport.h>
>>>> #include <asm/hap.h>
>>>> #include <asm/apic.h>
>>>> #include <asm/debugger.h>
>>>> @@ -2065,6 +2066,38 @@ svm_vmexit_do_vmsave(struct vmcb_struct *vmcb,
>>>> return;
>>>> }
>>>> +static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
>>>> + struct vcpu *v)
>>>> +{
>>>> + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>>> + unsigned long inst_len;
>>>> + unsigned long inst_addr = svm_rip2pointer(v);
>>>> + int rc;
>>>> + static const enum instruction_index list[] = {
>>>> + INSTR_INL_DX, INSTR_INB_DX, INSTR_OUTL_DX, INSTR_OUTB_DX
>>>> + };
>>>> +
>>>> + inst_len = __get_instruction_length_from_list(
>>>> + v, list, ARRAY_SIZE(list), 0);
>>> I should have asked earlier but I don't think I understand why the
>>> last argument here is 0 (and therefore why you have this last argument
>>> at all).
>>>
>>> Because whether or not you are warning in
>>> __get_instruction_length_from_list() it will still return 0. And that,
>>> in turn, will cause vmport_gp_check() to return an error. And then you
>>> will print another warning in VMPORT_LOG. So there is a warning anyway.
>>>
>> A key part that you appear to have missed is that
>> __get_instruction_length_from_list() uses gdprintk(XENLOG_WARNING,...
>> but VMPORT_DBG_LOG is only available in debug=y builds. So the new
>> last argument is used to control this. Since this change includes
>> enabling #GP vmexits, it is now possible for ring 3 users to generate at
>> large volume of these which with gdprintk() can flood the console.
>
> Would it be possible to decide where and whether to print the warning
> inside __get_instruction_length_from_list() as opposed to passing a
> new parameter? E.g. if vmware_port_enabled is set and list includes
> IN/OUT and possibly something else?
>
It could be. However the test is complex and not something I am willing
to be part of this patch. So the only thing I have come up with is using
#ifndef NDEBUG around the gdprintk(). This leaves the
hvm_inject_hw_exception(TRAP_gp_fault, 0);
Which is not correct in this case. Now are far as I can tell, calling this
twice does the wrong thing for this case. I.E. turns it into a double
fault.
For #GP I need to call it with vmcb->exitinfo1 instead of 0.
(Note: v5 posted before I got this.)
I just spent some time checking and found out that even with the cpu
reporting:
(XEN) - Next-RIP Saved on #VMEXIT
svm_nextrip_insn_length(v) is 0.
(hyper-0-21-54:~>cat /proc/cpuinfo
processor : 0
vendor_id : AuthenticAMD
cpu family : 21
model : 2
model name : AMD Opteron(tm) Processor 4365 EE
stepping : 0
microcode : 0x600081c
...)
Which means to me that by using __get_instruction_length_from_list()
I am reading the instruction bytes 2 times. Once in
__get_instruction_length_from_list() and once in vmport_gp_check().
This is not too bad since the rate of these is low. But this does
suggest to
me that the change to using __get_instruction_length_from_list() was
not the best. Having a routine that both checks the instruction and
reports on which one was found would be much better.
So do I go with v5, or dropping the use of
__get_instruction_length_from_list()
in a v6? (The coding of the new routine will take some time.)
-Don Slutz
> -boris
>
>>
>>
>>> Second, since this handler appears to be handling #GP only for VMware
>>> guest we should make sure that it is not executed for any other guest.
>>> You do now condition intercept got #GP for such guests only but I
>>> still think having a check here is worth doing. Maybe a BUG() or
>>> ASSERT()?
>>>
>>> The same comments are applicable to VMX code, I suspect.
>>>
>> I will change the check in vmport_gp_check on is_vmware_port_enabled
>> into an ASSERT() so both SVM and VMX will be covered.
>>
>>>> +
>>>> + rc = vmport_gp_check(regs, v, inst_len, inst_addr,
>>>> + vmcb->exitinfo1, vmcb->exitinfo2);
>>>> + if ( !rc )
>>>> + __update_guest_eip(regs, inst_len);
>>>> + else
>>>> + {
>>>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_UNKNOWN,
>>>> + "gp: rc=%d ei1=0x%lx ei2=0x%lx ip=%"PRIx64
>>>> + " (0x%lx,%ld) ax=%"PRIx64" bx=%"PRIx64"
>>>> cx=%"PRIx64
>>>> + " dx=%"PRIx64" si=%"PRIx64" di=%"PRIx64, rc,
>>>> + (unsigned long)vmcb->exitinfo1,
>>>> + (unsigned long)vmcb->exitinfo2, regs->rip,
>>>> inst_addr,
>>>> + inst_len, regs->rax, regs->rbx, regs->rcx,
>>>> regs->rdx,
>>>> + regs->rsi, regs->rdi);
>>>> + hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
>>>> + }
>>>> +}
>>>> +
>>> .
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-17 18:23 ` Slutz, Donald Christopher
@ 2014-09-18 9:14 ` Jan Beulich
2014-09-19 12:48 ` Slutz, Donald Christopher
2014-09-18 22:53 ` Boris Ostrovsky
1 sibling, 1 reply; 51+ messages in thread
From: Jan Beulich @ 2014-09-18 9:14 UTC (permalink / raw)
To: Donald Christopher Slutz
Cc: Jun Nakajima, Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
xen-devel, Eddie Dong, AravindGopalakrishnan,
Suravee Suthikulpanit, Boris Ostrovsky
>>> On 17.09.14 at 20:23, <dslutz@verizon.com> wrote:
> Which means to me that by using __get_instruction_length_from_list()
> I am reading the instruction bytes 2 times. Once in
> __get_instruction_length_from_list() and once in vmport_gp_check().
> This is not too bad since the rate of these is low. But this does
> suggest to
> me that the change to using __get_instruction_length_from_list() was
> not the best. Having a routine that both checks the instruction and
> reports on which one was found would be much better.
Just to clarify - for security reasons we should avoid reading
instructions more than once, or be very certain that the second
reader doesn't make any implications on eventual checks the first
one may have done.
Jan
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-17 18:23 ` Slutz, Donald Christopher
2014-09-18 9:14 ` Jan Beulich
@ 2014-09-18 22:53 ` Boris Ostrovsky
2014-09-19 13:24 ` Slutz, Donald Christopher
1 sibling, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-18 22:53 UTC (permalink / raw)
To: Slutz, Donald Christopher
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
George Dunlap, Eddie Dong, Tim Deegan, xen-devel, Jan Beulich,
Aravind Gopalakrishnan, Jun Nakajima, Andrew Cooper,
Suravee Suthikulpanit, Ian Jackson
On 09/17/2014 02:23 PM, Slutz, Donald Christopher wrote:
> On 09/17/14 11:56, Boris Ostrovsky wrote:
>> On 09/16/2014 08:08 AM, Slutz, Donald Christopher wrote:
>>> On 09/12/14 09:08, Boris Ostrovsky wrote:
>>>> On 09/11/2014 02:36 PM, Don Slutz wrote:
>>>>> int __get_instruction_length_from_list(struct vcpu *v,
>>>>> - const enum instruction_index *list, unsigned int list_count)
>>>>> + const enum instruction_index
>>>>> *list,
>>>>> + unsigned int list_count,
>>>>> + bool_t err_rpt)
>>>>> {
>>>>> struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>>>> unsigned int i, j, inst_len = 0;
>>>>> @@ -211,10 +222,13 @@ int __get_instruction_length_from_list(struct
>>>>> vcpu *v,
>>>>> mismatch: ;
>>>>> }
>>>>> - gdprintk(XENLOG_WARNING,
>>>>> - "%s: Mismatch between expected and actual instruction
>>>>> bytes: "
>>>>> - "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>>>>> - hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>>>> + if ( err_rpt )
>>>>> + {
>>>>> + gdprintk(XENLOG_WARNING,
>>>>> + "%s: Mismatch between expected and actual
>>>>> instruction bytes: "
>>>>> + "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>>>>> + hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>>>> + }
>>>>> return 0;
>>>>> done:
>>>>> diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
>>>>> index b5188e6..9e14d2a 100644
>>>>> --- a/xen/arch/x86/hvm/svm/svm.c
>>>>> +++ b/xen/arch/x86/hvm/svm/svm.c
>>>>> @@ -59,6 +59,7 @@
>>>>> #include <public/sched.h>
>>>>> #include <asm/hvm/vpt.h>
>>>>> #include <asm/hvm/trace.h>
>>>>> +#include <asm/hvm/vmport.h>
>>>>> #include <asm/hap.h>
>>>>> #include <asm/apic.h>
>>>>> #include <asm/debugger.h>
>>>>> @@ -2065,6 +2066,38 @@ svm_vmexit_do_vmsave(struct vmcb_struct *vmcb,
>>>>> return;
>>>>> }
>>>>> +static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
>>>>> + struct vcpu *v)
>>>>> +{
>>>>> + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>>>> + unsigned long inst_len;
>>>>> + unsigned long inst_addr = svm_rip2pointer(v);
>>>>> + int rc;
>>>>> + static const enum instruction_index list[] = {
>>>>> + INSTR_INL_DX, INSTR_INB_DX, INSTR_OUTL_DX, INSTR_OUTB_DX
>>>>> + };
>>>>> +
>>>>> + inst_len = __get_instruction_length_from_list(
>>>>> + v, list, ARRAY_SIZE(list), 0);
>>>> I should have asked earlier but I don't think I understand why the
>>>> last argument here is 0 (and therefore why you have this last argument
>>>> at all).
>>>>
>>>> Because whether or not you are warning in
>>>> __get_instruction_length_from_list() it will still return 0. And that,
>>>> in turn, will cause vmport_gp_check() to return an error. And then you
>>>> will print another warning in VMPORT_LOG. So there is a warning anyway.
>>>>
>>> A key part that you appear to have missed is that
>>> __get_instruction_length_from_list() uses gdprintk(XENLOG_WARNING,...
>>> but VMPORT_DBG_LOG is only available in debug=y builds. So the new
>>> last argument is used to control this. Since this change includes
>>> enabling #GP vmexits, it is now possible for ring 3 users to generate at
>>> large volume of these which with gdprintk() can flood the console.
>> Would it be possible to decide where and whether to print the warning
>> inside __get_instruction_length_from_list() as opposed to passing a
>> new parameter? E.g. if vmware_port_enabled is set and list includes
>> IN/OUT and possibly something else?
>>
> It could be. However the test is complex and not something I am willing
> to be part of this patch. So the only thing I have come up with is using
> #ifndef NDEBUG around the gdprintk(). This leaves the
>
>
> hvm_inject_hw_exception(TRAP_gp_fault, 0);
>
> Which is not correct in this case. Now are far as I can tell, calling this
> twice does the wrong thing for this case. I.E. turns it into a double
> fault.
> For #GP I need to call it with vmcb->exitinfo1 instead of 0.
>
> (Note: v5 posted before I got this.)
>
> I just spent some time checking and found out that even with the cpu
> reporting:
>
> (XEN) - Next-RIP Saved on #VMEXIT
>
> svm_nextrip_insn_length(v) is 0.
As I see it, this could happen for one of three reasons:
* !cpu_has_svm_nrips which can't be the case since (1) family 21
supports it and (2) you actually see in the log that it does have it
* next RIP is before current RIP which I think can't be the case neither
because we are not looking at branch instruction or something like that.
* nextrip == rip. Which I don't see how it can be true.
Can you check why svm_nextrip_insn_length(v) is 0?
But regardless of that, how do you expect your code to work on CPUs that
don't support NRIP? On those processors you *will* be decoding the
instruction twice.
>
> (hyper-0-21-54:~>cat /proc/cpuinfo
> processor : 0
> vendor_id : AuthenticAMD
> cpu family : 21
> model : 2
> model name : AMD Opteron(tm) Processor 4365 EE
> stepping : 0
> microcode : 0x600081c
> ...)
>
> Which means to me that by using __get_instruction_length_from_list()
> I am reading the instruction bytes 2 times. Once in
> __get_instruction_length_from_list() and once in vmport_gp_check().
> This is not too bad since the rate of these is low. But this does
> suggest to
> me that the change to using __get_instruction_length_from_list() was
> not the best. Having a routine that both checks the instruction and
> reports on which one was found would be much better.
>
> So do I go with v5, or dropping the use of
> __get_instruction_length_from_list()
> in a v6? (The coding of the new routine will take some time.)
Given what Jan said, it sounds like v5 is not going to work since you'd
be decoding instruction twice, right?
-boris
>
> -Don Slutz
>
>
>> -boris
>>
>>>
>>>> Second, since this handler appears to be handling #GP only for VMware
>>>> guest we should make sure that it is not executed for any other guest.
>>>> You do now condition intercept got #GP for such guests only but I
>>>> still think having a check here is worth doing. Maybe a BUG() or
>>>> ASSERT()?
>>>>
>>>> The same comments are applicable to VMX code, I suspect.
>>>>
>>> I will change the check in vmport_gp_check on is_vmware_port_enabled
>>> into an ASSERT() so both SVM and VMX will be covered.
>>>
>>>>> +
>>>>> + rc = vmport_gp_check(regs, v, inst_len, inst_addr,
>>>>> + vmcb->exitinfo1, vmcb->exitinfo2);
>>>>> + if ( !rc )
>>>>> + __update_guest_eip(regs, inst_len);
>>>>> + else
>>>>> + {
>>>>> + VMPORT_DBG_LOG(VMPORT_LOG_GP_UNKNOWN,
>>>>> + "gp: rc=%d ei1=0x%lx ei2=0x%lx ip=%"PRIx64
>>>>> + " (0x%lx,%ld) ax=%"PRIx64" bx=%"PRIx64"
>>>>> cx=%"PRIx64
>>>>> + " dx=%"PRIx64" si=%"PRIx64" di=%"PRIx64, rc,
>>>>> + (unsigned long)vmcb->exitinfo1,
>>>>> + (unsigned long)vmcb->exitinfo2, regs->rip,
>>>>> inst_addr,
>>>>> + inst_len, regs->rax, regs->rbx, regs->rcx,
>>>>> regs->rdx,
>>>>> + regs->rsi, regs->rdi);
>>>>> + hvm_inject_hw_exception(TRAP_gp_fault, regs->error_code);
>>>>> + }
>>>>> +}
>>>>> +
>>>> .
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-18 9:14 ` Jan Beulich
@ 2014-09-19 12:48 ` Slutz, Donald Christopher
0 siblings, 0 replies; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-19 12:48 UTC (permalink / raw)
To: Jan Beulich
Cc: Jun Nakajima, Tim Deegan, Kevin Tian, Keir Fraser, Ian Campbell,
Stefano Stabellini, George Dunlap, Andrew Cooper, Ian Jackson,
xen-devel, Eddie Dong, AravindGopalakrishnan,
Suravee Suthikulpanit, Boris Ostrovsky
On 09/18/14 05:14, Jan Beulich wrote:
>>>> On 17.09.14 at 20:23, <dslutz@verizon.com> wrote:
>> Which means to me that by using __get_instruction_length_from_list()
>> I am reading the instruction bytes 2 times. Once in
>> __get_instruction_length_from_list() and once in vmport_gp_check().
>> This is not too bad since the rate of these is low. But this does
>> suggest to
>> me that the change to using __get_instruction_length_from_list() was
>> not the best. Having a routine that both checks the instruction and
>> reports on which one was found would be much better.
> Just to clarify - for security reasons we should avoid reading
> instructions more than once, or be very certain that the second
> reader doesn't make any implications on eventual checks the first
> one may have done.
Assuming I am reading this correctly, the current code passes this
statement.
I am looking into a way to skip this 2nd read & decode.
-Don Slutz
> Jan
>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-18 22:53 ` Boris Ostrovsky
@ 2014-09-19 13:24 ` Slutz, Donald Christopher
2014-09-19 15:50 ` Boris Ostrovsky
0 siblings, 1 reply; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-19 13:24 UTC (permalink / raw)
To: Boris Ostrovsky
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
George Dunlap, Eddie Dong, Tim Deegan, xen-devel, Jan Beulich,
Aravind Gopalakrishnan, Jun Nakajima, Andrew Cooper,
Suravee Suthikulpanit, Ian Jackson
On 09/18/14 18:53, Boris Ostrovsky wrote:
> On 09/17/2014 02:23 PM, Slutz, Donald Christopher wrote:
>> On 09/17/14 11:56, Boris Ostrovsky wrote:
>>> On 09/16/2014 08:08 AM, Slutz, Donald Christopher wrote:
>>>> On 09/12/14 09:08, Boris Ostrovsky wrote:
>>>>> On 09/11/2014 02:36 PM, Don Slutz wrote:
>>>>>> int __get_instruction_length_from_list(struct vcpu *v,
>>>>>> - const enum instruction_index *list, unsigned int
>>>>>> list_count)
>>>>>> + const enum instruction_index
>>>>>> *list,
>>>>>> + unsigned int list_count,
>>>>>> + bool_t err_rpt)
>>>>>> {
>>>>>> struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>>>>> unsigned int i, j, inst_len = 0;
>>>>>> @@ -211,10 +222,13 @@ int __get_instruction_length_from_list(struct
>>>>>> vcpu *v,
>>>>>> mismatch: ;
>>>>>> }
>>>>>> - gdprintk(XENLOG_WARNING,
>>>>>> - "%s: Mismatch between expected and actual instruction
>>>>>> bytes: "
>>>>>> - "eip = %lx\n", __func__, (unsigned long)vmcb->rip);
>>>>>> - hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>>>>> + if ( err_rpt )
>>>>>> + {
>>>>>> + gdprintk(XENLOG_WARNING,
>>>>>> + "%s: Mismatch between expected and actual
>>>>>> instruction bytes: "
>>>>>> + "eip = %lx\n", __func__, (unsigned
>>>>>> long)vmcb->rip);
>>>>>> + hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>>>>> + }
>>>>>> return 0;
>>>>>> done:
>>>>>> diff --git a/xen/arch/x86/hvm/svm/svm.c b/xen/arch/x86/hvm/svm/svm.c
>>>>>> index b5188e6..9e14d2a 100644
>>>>>> --- a/xen/arch/x86/hvm/svm/svm.c
>>>>>> +++ b/xen/arch/x86/hvm/svm/svm.c
>>>>>> @@ -59,6 +59,7 @@
>>>>>> #include <public/sched.h>
>>>>>> #include <asm/hvm/vpt.h>
>>>>>> #include <asm/hvm/trace.h>
>>>>>> +#include <asm/hvm/vmport.h>
>>>>>> #include <asm/hap.h>
>>>>>> #include <asm/apic.h>
>>>>>> #include <asm/debugger.h>
>>>>>> @@ -2065,6 +2066,38 @@ svm_vmexit_do_vmsave(struct vmcb_struct
>>>>>> *vmcb,
>>>>>> return;
>>>>>> }
>>>>>> +static void svm_vmexit_gp_intercept(struct cpu_user_regs *regs,
>>>>>> + struct vcpu *v)
>>>>>> +{
>>>>>> + struct vmcb_struct *vmcb = v->arch.hvm_svm.vmcb;
>>>>>> + unsigned long inst_len;
>>>>>> + unsigned long inst_addr = svm_rip2pointer(v);
>>>>>> + int rc;
>>>>>> + static const enum instruction_index list[] = {
>>>>>> + INSTR_INL_DX, INSTR_INB_DX, INSTR_OUTL_DX, INSTR_OUTB_DX
>>>>>> + };
>>>>>> +
>>>>>> + inst_len = __get_instruction_length_from_list(
>>>>>> + v, list, ARRAY_SIZE(list), 0);
>>>>> I should have asked earlier but I don't think I understand why the
>>>>> last argument here is 0 (and therefore why you have this last
>>>>> argument
>>>>> at all).
>>>>>
>>>>> Because whether or not you are warning in
>>>>> __get_instruction_length_from_list() it will still return 0. And
>>>>> that,
>>>>> in turn, will cause vmport_gp_check() to return an error. And then
>>>>> you
>>>>> will print another warning in VMPORT_LOG. So there is a warning
>>>>> anyway.
>>>>>
>>>> A key part that you appear to have missed is that
>>>> __get_instruction_length_from_list() uses gdprintk(XENLOG_WARNING,...
>>>> but VMPORT_DBG_LOG is only available in debug=y builds. So the new
>>>> last argument is used to control this. Since this change includes
>>>> enabling #GP vmexits, it is now possible for ring 3 users to
>>>> generate at
>>>> large volume of these which with gdprintk() can flood the console.
>>> Would it be possible to decide where and whether to print the warning
>>> inside __get_instruction_length_from_list() as opposed to passing a
>>> new parameter? E.g. if vmware_port_enabled is set and list includes
>>> IN/OUT and possibly something else?
>>>
>> It could be. However the test is complex and not something I am willing
>> to be part of this patch. So the only thing I have come up with is
>> using
>> #ifndef NDEBUG around the gdprintk(). This leaves the
>>
>>
>> hvm_inject_hw_exception(TRAP_gp_fault, 0);
>>
>> Which is not correct in this case. Now are far as I can tell,
>> calling this
>> twice does the wrong thing for this case. I.E. turns it into a double
>> fault.
>> For #GP I need to call it with vmcb->exitinfo1 instead of 0.
>>
>> (Note: v5 posted before I got this.)
>>
>> I just spent some time checking and found out that even with the cpu
>> reporting:
>>
>> (XEN) - Next-RIP Saved on #VMEXIT
>>
>> svm_nextrip_insn_length(v) is 0.
>
> As I see it, this could happen for one of three reasons:
> * !cpu_has_svm_nrips which can't be the case since (1) family 21
> supports it and (2) you actually see in the log that it does have it
> * next RIP is before current RIP which I think can't be the case
> neither because we are not looking at branch instruction or something
> like that.
> * nextrip == rip. Which I don't see how it can be true.
>
> Can you check why svm_nextrip_insn_length(v) is 0?
>
What I have found out is that vmcb->nextrip is 0 in my testing. So
next RIP is "before current RIP" by a very large amount.
> But regardless of that, how do you expect your code to work on CPUs
> that don't support NRIP? On those processors you *will* be decoding
> the instruction twice.
>
The code "works" because the only info passed between the 2 decodes is
the instruction length. This is used to limit the amount of the 2nd read.
And because svm_nextrip_insn_length(v) is 0, all the testing I have done
on the AMD server has been doing the decode of the instruction twice.
If another CPU is changing the instructions as I read them (which is the
security issue as I understand it), all I see happing is that "wrong"
direction
or size of request could happen, or it is a vmport request or not. All of
which either report a #GP or do the vmport action. Since you can do the
vmport action without changing the instruction bytes, I do not see how
the double decode opens any security issue.
I am not trying to say this is good. And as I replied to Jan, I am
looking into
a way to only do the single decode. The simplest of these is to just not
use __get_instruction_length_from_list() and just state that on AMD the
instruction length is 2. This is safe because I am only using this to
decide
much many bytes on the page to read. The 2nd page read read depends
on finding a 0x66 prefix byte. I am just noticing that this also has the
side
effect of not injecting a #PF if the instruction is no longer readable
(a side
effect of using fetch() which uses hvm_fetch_from_guest_virt()).
>>
>> (hyper-0-21-54:~>cat /proc/cpuinfo
>> processor : 0
>> vendor_id : AuthenticAMD
>> cpu family : 21
>> model : 2
>> model name : AMD Opteron(tm) Processor 4365 EE
>> stepping : 0
>> microcode : 0x600081c
>> ...)
>>
>> Which means to me that by using __get_instruction_length_from_list()
>> I am reading the instruction bytes 2 times. Once in
>> __get_instruction_length_from_list() and once in vmport_gp_check().
>> This is not too bad since the rate of these is low. But this does
>> suggest to
>> me that the change to using __get_instruction_length_from_list() was
>> not the best. Having a routine that both checks the instruction and
>> reports on which one was found would be much better.
>>
>> So do I go with v5, or dropping the use of
>> __get_instruction_length_from_list()
>> in a v6? (The coding of the new routine will take some time.)
>
> Given what Jan said, it sounds like v5 is not going to work since
> you'd be decoding instruction twice, right?
>
My take is that v5 "works" because the 2nd decode does not use any info
from the 1st decode.
Sigh, I have to take that back. The check "fetch_len != inst_len" in:
if ( bytes[0] == 0x66 ) /* operand size prefix */
{
byte_cnt = 2;
i = 1;
if ( fetch_len != inst_len )
{
does violate Jan's statement. The simple change of inst_len to 2 would
fix it. As I replied to Jan, I am looking into how to not read and decode
more then one time.
-Don Slutz
> -boris
>
>>
>> -Don Slutz
>>
>>
>>> -boris
>>>
>>>>
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-19 13:24 ` Slutz, Donald Christopher
@ 2014-09-19 15:50 ` Boris Ostrovsky
2014-09-19 17:00 ` Slutz, Donald Christopher
0 siblings, 1 reply; 51+ messages in thread
From: Boris Ostrovsky @ 2014-09-19 15:50 UTC (permalink / raw)
To: Slutz, Donald Christopher
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
George Dunlap, Eddie Dong, Tim Deegan, xen-devel, Jan Beulich,
Aravind Gopalakrishnan, Jun Nakajima, Andrew Cooper,
Suravee Suthikulpanit, Ian Jackson
On 09/19/2014 09:24 AM, Slutz, Donald Christopher wrote:
> On 09/18/14 18:53, Boris Ostrovsky wrote:
>> On 09/17/2014 02:23 PM, Slutz, Donald Christopher wrote:
>>>
>>> I just spent some time checking and found out that even with the cpu
>>> reporting:
>>>
>>> (XEN) - Next-RIP Saved on #VMEXIT
>>>
>>> svm_nextrip_insn_length(v) is 0.
>> As I see it, this could happen for one of three reasons:
>> * !cpu_has_svm_nrips which can't be the case since (1) family 21
>> supports it and (2) you actually see in the log that it does have it
>> * next RIP is before current RIP which I think can't be the case
>> neither because we are not looking at branch instruction or something
>> like that.
>> * nextrip == rip. Which I don't see how it can be true.
>>
>> Can you check why svm_nextrip_insn_length(v) is 0?
>>
> What I have found out is that vmcb->nextrip is 0 in my testing. So
> next RIP is "before current RIP" by a very large amount.
Ah, I just realized --- you are in VMEXIT because of #GP intercept, not
because of IOIO intercept. And during #GP NRIP does not get updated (it
is set to zero).
Which means that you will always be doing decoding when you call
__get_instruction_length_from_list(), regardless of NRIP support.
>> But regardless of that, how do you expect your code to work on CPUs
>> that don't support NRIP? On those processors you *will* be decoding
>> the instruction twice.
>>
> The code "works" because the only info passed between the 2 decodes is
> the instruction length. This is used to limit the amount of the 2nd read.
>
> And because svm_nextrip_insn_length(v) is 0, all the testing I have done
> on the AMD server has been doing the decode of the instruction twice.
>
> If another CPU is changing the instructions as I read them (which is the
> security issue as I understand it), all I see happing is that "wrong"
> direction
> or size of request could happen, or it is a vmport request or not. All of
> which either report a #GP or do the vmport action. Since you can do the
> vmport action without changing the instruction bytes, I do not see how
> the double decode opens any security issue.
>
> I am not trying to say this is good. And as I replied to Jan, I am
> looking into
> a way to only do the single decode. The simplest of these is to just not
> use __get_instruction_length_from_list() and just state that on AMD the
> instruction length is 2. This is safe because I am only using this to
> decide
> much many bytes on the page to read. The 2nd page read read depends
> on finding a 0x66 prefix byte.
Can you make this statement (about instruction length being 2) for both
AMD and Intel and then possibly move some code into common HVM code?
Since we are intercepting #GP on both architectures only for VMware case
(right?) it seems that you can just say -- "let's look at the next 2
bytes and confirm that they are IN/OUT (with appropriate arguments)".
-boris
> I am just noticing that this also has the
> side
> effect of not injecting a #PF if the instruction is no longer readable
> (a side
> effect of using fetch() which uses hvm_fetch_from_guest_virt()).
^ permalink raw reply [flat|nested] 51+ messages in thread
* Re: [PATCH v4 04/16] xen: Add is_vmware_port_enabled
2014-09-19 15:50 ` Boris Ostrovsky
@ 2014-09-19 17:00 ` Slutz, Donald Christopher
0 siblings, 0 replies; 51+ messages in thread
From: Slutz, Donald Christopher @ 2014-09-19 17:00 UTC (permalink / raw)
To: Boris Ostrovsky
Cc: Kevin Tian, Keir Fraser, Ian Campbell, Stefano Stabellini,
George Dunlap, Eddie Dong, Tim Deegan, xen-devel, Jan Beulich,
Aravind Gopalakrishnan, Jun Nakajima, Andrew Cooper,
Suravee Suthikulpanit, Ian Jackson
On 09/19/14 11:50, Boris Ostrovsky wrote:
> On 09/19/2014 09:24 AM, Slutz, Donald Christopher wrote:
>> On 09/18/14 18:53, Boris Ostrovsky wrote:
>>> On 09/17/2014 02:23 PM, Slutz, Donald Christopher wrote:
>>>>
>>>> I just spent some time checking and found out that even with the cpu
>>>> reporting:
>>>>
>>>> (XEN) - Next-RIP Saved on #VMEXIT
>>>>
>>>> svm_nextrip_insn_length(v) is 0.
>>> As I see it, this could happen for one of three reasons:
>>> * !cpu_has_svm_nrips which can't be the case since (1) family 21
>>> supports it and (2) you actually see in the log that it does have it
>>> * next RIP is before current RIP which I think can't be the case
>>> neither because we are not looking at branch instruction or something
>>> like that.
>>> * nextrip == rip. Which I don't see how it can be true.
>>>
>>> Can you check why svm_nextrip_insn_length(v) is 0?
>>>
>> What I have found out is that vmcb->nextrip is 0 in my testing. So
>> next RIP is "before current RIP" by a very large amount.
>
>
> Ah, I just realized --- you are in VMEXIT because of #GP intercept,
> not because of IOIO intercept. And during #GP NRIP does not get
> updated (it is set to zero).
>
> Which means that you will always be doing decoding when you call
> __get_instruction_length_from_list(), regardless of NRIP support.
>
>
>>> But regardless of that, how do you expect your code to work on CPUs
>>> that don't support NRIP? On those processors you *will* be decoding
>>> the instruction twice.
>>>
>> The code "works" because the only info passed between the 2 decodes is
>> the instruction length. This is used to limit the amount of the 2nd
>> read.
>>
>> And because svm_nextrip_insn_length(v) is 0, all the testing I have done
>> on the AMD server has been doing the decode of the instruction twice.
>>
>> If another CPU is changing the instructions as I read them (which is the
>> security issue as I understand it), all I see happing is that "wrong"
>> direction
>> or size of request could happen, or it is a vmport request or not.
>> All of
>> which either report a #GP or do the vmport action. Since you can do the
>> vmport action without changing the instruction bytes, I do not see how
>> the double decode opens any security issue.
>>
>> I am not trying to say this is good. And as I replied to Jan, I am
>> looking into
>> a way to only do the single decode. The simplest of these is to just
>> not
>> use __get_instruction_length_from_list() and just state that on AMD the
>> instruction length is 2. This is safe because I am only using this to
>> decide
>> much many bytes on the page to read. The 2nd page read read depends
>> on finding a 0x66 prefix byte.
>
> Can you make this statement (about instruction length being 2) for
> both AMD and Intel and then possibly move some code into common HVM code?
>
I could make this statement. However there is no code movement because
of this. And Intel uses update_guest_eip() vs AMD's
__update_guest_eip(regs, inst_len). Since update_guest_eip() uses
the same field (VM_EXIT_INSTRUCTION_LEN). I did not use
get_instruction_length() because I am not 100% sure that it cannot return
a 0, and get_instruction_length() would BUG in that case. However when
update_guest_eip() is called, I know that get_instruction_length() will not
BUG.
My reading of the Intel Software Developer's Manual says that the byte
sequence "0x66 0x66 0xed" or "0x66 0x66 0x66 0xed" are valid, just
a waste of code space. I see no reason to allow this in my decode for
VMware IN/OUT. But the get_instruction_length() on Intel would
return 3 or 4 for these. I plan on adjusting my unit test code to see what
VMware does in this case.
From Intel Software Developer's Manual 2A section 2.1 I should use 5 not 2
if I am supporting this useless encoding.
> Since we are intercepting #GP on both architectures only for VMware
> case (right?) it seems that you can just say -- "let's look at the
> next 2 bytes and confirm that they are IN/OUT (with appropriate
> arguments)".
>
Almost. The #GP that I accept is only for VMware case. When enabled you
do get all #GPs, not just the VMware ones. And your statement is a good
description of vmport_gp_check.
-Don Slutz
> -boris
>
>> I am just noticing that this also has the
>> side
>> effect of not injecting a #PF if the instruction is no longer readable
>> (a side
>> effect of using fetch() which uses hvm_fetch_from_guest_virt()).
>
^ permalink raw reply [flat|nested] 51+ messages in thread
end of thread, other threads:[~2014-09-19 17:00 UTC | newest]
Thread overview: 51+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2014-09-11 18:36 [PATCH v4 00/16] Xen VMware tools support Don Slutz
2014-09-11 18:36 ` [PATCH v4 01/16] xen: Add support for VMware cpuid leaves Don Slutz
2014-09-11 19:49 ` Andrew Cooper
2014-09-12 9:49 ` Jan Beulich
2014-09-12 17:46 ` Don Slutz
2014-09-15 7:42 ` Jan Beulich
2014-09-17 15:41 ` Don Slutz
2014-09-12 21:26 ` Don Slutz
2014-09-12 12:37 ` Boris Ostrovsky
2014-09-12 17:50 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 02/16] tools: Add vmware_hw support Don Slutz
2014-09-11 21:09 ` Andrew Cooper
2014-09-16 16:20 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 03/16] vmware: Add VMware provided include files Don Slutz
2014-09-11 18:36 ` [PATCH v4 04/16] xen: Add is_vmware_port_enabled Don Slutz
2014-09-11 21:22 ` Andrew Cooper
2014-09-16 16:20 ` Don Slutz
2014-09-12 13:08 ` Boris Ostrovsky
2014-09-16 12:08 ` Slutz, Donald Christopher
2014-09-17 15:56 ` Boris Ostrovsky
2014-09-17 18:23 ` Slutz, Donald Christopher
2014-09-18 9:14 ` Jan Beulich
2014-09-19 12:48 ` Slutz, Donald Christopher
2014-09-18 22:53 ` Boris Ostrovsky
2014-09-19 13:24 ` Slutz, Donald Christopher
2014-09-19 15:50 ` Boris Ostrovsky
2014-09-19 17:00 ` Slutz, Donald Christopher
2014-09-11 18:36 ` [PATCH v4 05/16] tools: Add vmware_port support Don Slutz
2014-09-11 18:36 ` [PATCH v4 06/16] xen: Convert vmware_port to xentrace usage Don Slutz
2014-09-12 13:10 ` Boris Ostrovsky
2014-09-12 23:57 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 07/16] tools: " Don Slutz
2014-09-12 13:15 ` Boris Ostrovsky
2014-09-13 0:01 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 08/16] xen: Add limited support of VMware's hyper-call rpc Don Slutz
2014-09-12 13:37 ` Boris Ostrovsky
2014-09-12 14:27 ` Jan Beulich
2014-09-16 12:38 ` Slutz, Donald Christopher
2014-09-16 12:46 ` Jan Beulich
2014-09-16 13:47 ` Slutz, Donald Christopher
2014-09-16 13:17 ` Slutz, Donald Christopher
2014-09-11 18:36 ` [PATCH v4 09/16] tools: " Don Slutz
2014-09-11 18:36 ` [PATCH v4 10/16] Add VMware tool's triggers Don Slutz
2014-09-11 18:36 ` [PATCH v4 11/16] Add live migration of VMware's hyper-call RPC Don Slutz
2014-09-12 13:54 ` Boris Ostrovsky
2014-09-16 15:48 ` Don Slutz
2014-09-11 18:36 ` [PATCH v4 12/16] Add dump of HVM_SAVE_CODE(VMPORT) to xen-hvmctx Don Slutz
2014-09-11 18:36 ` [PATCH v4 13/16] Add xen-hvm-param Don Slutz
2014-09-11 18:36 ` [PATCH v4 14/16] Add xen-vmware-guestinfo Don Slutz
2014-09-11 18:36 ` [PATCH v4 15/16] Add xen-list-vmware-guestinfo Don Slutz
2014-09-11 18:36 ` [PATCH v4 16/16] Add xen-hvm-send-trigger Don Slutz
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.