Argo: hypervisor-mediated interdomain communication
v7 cover Argo: hypervisor-mediated interdomain communication
Version seven of this series:
* Adds compat ABI in support of using handles
- hypercall table entries now use COMPAT_CALL
- affects boilerplate patch, so dropped the prior Ack there
- ARM hypercall table entry is actually added this time
- since using compat now, dropped use of always-64-bit pfn type
(note: this is first time adding such compat logic: feedback
appreciated)
* Applied list iteration macros to replace for loops
* Fixes for XSM patches identified in testing
* Reorder and reposition compat entries in xlat.lst
- added missing xen_argo_message_header entry
Christopher Clark (15):
argo: Introduce the Kconfig option to govern inclusion of Argo
argo: introduce the argo_op hypercall boilerplate
argo: define argo_dprintk for subsystem debugging
argo: init, destroy and soft-reset, with enable command line opt
errno: add POSIX error codes EMSGSIZE, ECONNREFUSED to the ABI xen/arm: introduce guest_handle_for_field()
argo: implement the register op
argo: implement the unregister op
argo: implement the sendv op; evtchn: expose send_guest_global_virq
argo: implement the notify op
xsm, argo: XSM control for argo register
xsm, argo: XSM control for argo message send operation xsm, argo: XSM control for any access to argo by a domain xsm, argo: notify: don't describe rings that cannot be sent to
MAINTAINERS: add new section for Argo and self as maintainer
MAINTAINERS | 7 + docs/misc/xen-command-line.pandoc | 20 + tools/flask/policy/modules/guest_features.te | 7 + xen/arch/arm/traps.c | 3 + xen/arch/x86/guest/hypercall_page.S | 2 +- xen/arch/x86/hvm/hypercall.c | 3 + xen/arch/x86/hypercall.c | 3 + xen/arch/x86/pv/hypercall.c | 3 + xen/common/Kconfig | 19 + xen/common/Makefile | 1 + xen/common/argo.c | 2427 ++++++++++++++++++++++++++ xen/common/domain.c | 9 + xen/common/event_channel.c | 2 +- xen/include/Makefile | 1 + xen/include/asm-arm/guest_access.h | 3 + xen/include/public/argo.h | 274 +++ xen/include/public/errno.h | 2 + xen/include/public/xen.h | 4 +- xen/include/xen/argo.h | 44 + xen/include/xen/event.h | 7 + xen/include/xen/hypercall.h | 18 + xen/include/xen/sched.h | 5 + xen/include/xlat.lst | 9 + xen/include/xsm/dummy.h | 25 + xen/include/xsm/xsm.h | 31 + xen/xsm/dummy.c | 6 + xen/xsm/flask/hooks.c | 41 +- xen/xsm/flask/include/avc.h | 4 +- xen/xsm/flask/policy/access_vectors | 16 + xen/xsm/flask/policy/security_classes | 1 + 30 files changed, 2988 insertions(+), 9 deletions(-) create mode 100644 xen/common/argo.c create mode 100644 xen/include/public/argo.h create mode 100644 xen/include/xen/argo.h
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 1/15 argo: Introduce the Kconfig option to govern inclusion of Argo
Defines CONFIG_ARGO when enabled. Default: disabled.
When the Kconfig option is enabled, the Argo hypercall implementation will be included, allowing use of the hypervisor-mediated interdomain communication mechanism.
Argo is implemented for x86 and ARM hardware platforms.
Availability of the option depends on EXPERT and Argo is currently an experimental feature.
---
v3 added Jan's Ack
v2 #01 feedback, Jan: replace def_bool/prompt with bool
v1 #02 feedback, Jan: default Kconfig off, use EXPERT, fix whitespace
xen/common/Kconfig | 19 +++++++++++++++++++ 1 file changed, 19 insertions(+)
xen/common/Kconfig +19 -0
@@ -202,6 +202,25 @@ config LATE_HWDOM If unsure, say N. +config ARGO + bool "Argo: hypervisor-mediated interdomain communication" if EXPERT = "y" + ---help--- + Enables a hypercall for domains to ask the hypervisor to perform + data transfer of messages between domains. + + This allows communication channels to be established that do not + require any shared memory between domains; the hypervisor is the + entity that each domain interacts with. The hypervisor is able to + enforce Mandatory Access Control policy over the communication. + + If XSM_FLASK is enabled, XSM policy can govern which domains may + communicate via the Argo system. + + This feature does nothing if the "argo" boot parameter is not present. + Argo is disabled at runtime by default. + + If unsure, say N. + menu "Schedulers" visible if EXPERT = "y"
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 2/15 argo: introduce the argo_op hypercall boilerplate
In reply to Christopher ClarkPresence is gated upon CONFIG_ARGO.
Registers the hypercall previously reserved for this. Takes 5 arguments, does nothing and returns -ENOSYS.
Implementation will provide a compat ABI so COMPAT_CALL is the macro used in the hypercall tables.
---
v6 dropped Jan Acked-by due to change of implementation and commit msg v6 switched to COMPAT_CALL and provides compat_argo_op
v6 feedback #3 Julien: add argo_op to the ARM hypercall table v2 Copyright line: add 2019
v2 feedback #3 Jan: drop "message" from argo_message_op
v2 feedback #3 Jan: add Acked-by
v1 feedback #15 Jan: handle upper-halves of hypercall args v1 feedback #15 Jan: use unsigned where negative values impossible
xen/arch/arm/traps.c | 3 +++ xen/arch/x86/guest/hypercall_page.S | 2 +- xen/arch/x86/hvm/hypercall.c | 3 +++ xen/arch/x86/hypercall.c | 3 +++ xen/arch/x86/pv/hypercall.c | 3 +++ xen/common/Makefile | 1 + xen/common/argo.c | 38 +++++++++++++++++++++++++++++++++++++ xen/include/public/xen.h | 2 +- xen/include/xen/hypercall.h | 18 ++++++++++++++++++ 9 files changed, 71 insertions(+), 2 deletions(-) create mode 100644 xen/common/argo.c
xen/arch/arm/traps.c +3 -0
@@ -1397,6 +1397,9 @@ static arm_hypercall_t arm_hypercall_table[] = {HYPERCALL (platform_op,1 ),HYPERCALL_ARM (vcpu_op,3 ),HYPERCALL (vm_assist,2 ), +#ifdef CONFIG_ARGO + HYPERCALL(argo_op, 5), +#endif };#ifndef NDEBUG
xen/arch/x86/guest/hypercall_page.S +1 -1
@@ -59,7 +59,7 @@ DECLARE_HYPERCALL(sysctl) DECLARE_HYPERCALL(domctl) DECLARE_HYPERCALL(kexec_op) DECLARE_HYPERCALL(tmem_op) -DECLARE_HYPERCALL(xc_reserved_op) +DECLARE_HYPERCALL(argo_op) DECLARE_HYPERCALL(xenpmu_op) DECLARE_HYPERCALL(arch_0)
xen/arch/x86/hvm/hypercall.c +3 -0
@@ -134,6 +134,9 @@ static const hypercall_table_t hvm_hypercall_table[] = {#ifdef CONFIG_TMEM HYPERCALL(tmem_op),#endif +#ifdef CONFIG_ARGO + COMPAT_CALL(argo_op), +#endif COMPAT_CALL(platform_op),#ifdef CONFIG_PV COMPAT_CALL(mmuext_op),
xen/arch/x86/hypercall.c +3 -0
@@ -64,6 +64,9 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] =ARGS (domctl,1 ),ARGS (kexec_op,2 ),ARGS (tmem_op,1 ), +#ifdef CONFIG_ARGO +ARGS (argo_op,5 ), +#endifARGS (xenpmu_op,2 ), #ifdef CONFIG_HVM ARGS(hvm_op, 2),
xen/arch/x86/pv/hypercall.c +3 -0
@@ -77,6 +77,9 @@ const hypercall_table_t pv_hypercall_table[] = {#ifdef CONFIG_TMEM HYPERCALL(tmem_op),#endif +#ifdef CONFIG_ARGO + COMPAT_CALL(argo_op), +#endif HYPERCALL(xenpmu_op),#ifdef CONFIG_HVM HYPERCALL(hvm_op),
xen/common/Makefile +1 -0
@@ -1,3 +1,4 @@ +obj-$(CONFIG_ARGO) += argo.o obj-y += bitmap.o obj-y += bsearch.o obj-$(CONFIG_CORE_PARKING) += core_parking.o
xen/common/argo.c new file +38 -0
@@ -0,0 +1,38 @@ +/****************************************************************************** +* Argo : Hypervisor-Mediated data eXchange +* +* Derived from v4v, the version 2 of v2v. +* +* Copyright (c) 2010, Citrix Systems +* Copyright (c) 2018-2019 BAE Systems +* +* 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA +*/ + +#include <xen/errno.h> +#include <xen/guest_access.h> + +long +do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} + +#ifdef CONFIG_COMPAT +long +compat_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} +#endif
xen/include/public/xen.h +1 -1
@@ -118,7 +118,7 @@ DEFINE_XEN_GUEST_HANDLE(xen_ulong_t);#define __HYPERVISOR_domctl 36#define __HYPERVISOR_kexec_op 37#define __HYPERVISOR_tmem_op 38 -#define __HYPERVISOR_xc_reserved_op 39/* reserved for XenClient */ +#define __HYPERVISOR_argo_op 39#define __HYPERVISOR_xenpmu_op 40#define __HYPERVISOR_dm_op 41
xen/include/xen/hypercall.h +18 -0
@@ -136,6 +136,15 @@ do_tmem_op( XEN_GUEST_HANDLE_PARAM(tmem_op_t) uops);#endif +#ifdef CONFIG_ARGO +extern long do_argo_op ( +unsigned int cmd, +XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2, +unsigned long arg3, +unsigned long arg4); +#endif +extern long do_xenoprof_op (int op,XEN_GUEST_HANDLE_PARAM (void ) arg);
@@ -197,6 +206,15 @@ extern int compat_multicall(XEN_GUEST_HANDLE_PARAM (multicall_entry_compat_t) call_list, uint32_t nr_calls); +#ifdef CONFIG_ARGO +extern long compat_argo_op ( +unsigned int cmd, +XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2, +unsigned long arg3, +unsigned long arg4); +#endif +#include <compat/hvm/dm_op.h> extern int
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Quoted 6 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 05:28, <[email protected]> wrote:
Presence is gated upon
CONFIG_ARGO.Registers the hypercall previously reserved for this. Takes 5 arguments, does nothing and returns
-ENOSYS.
Quoted 6 lines
Implementation will provide a compat ABI so
COMPAT_CALLis the macro used in the hypercall tables.Signed-off-by Christopher Clark <[email protected]>---
v6 dropped Jan Acked-by due to change of implementation and commit msg
Feel free to re-instate.
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
In reply to Christopher ClarkQuoted 11 lines · Christopher Clark
Quoted 5 lines · Christopher Clark
Quoted 3 lines · Julien Grall
[...]
Sorry for noticing quite late in the process. Don't you need to add the hypercall inxen/arch/arm/traps.c?Adding this looked fine, so I've added the ARM hypercall table entry to this patch.
Julien, do you have any further feedback on the latest version of this patch?
Christopher
Quoted 204 lines · Christopher Clark
Presence is gated upon
CONFIG_ARGO.Registers the hypercall previously reserved for this. Takes 5 arguments, does nothing and returns
-ENOSYS.Implementation will provide a compat ABI so
COMPAT_CALLis the macro used in the hypercall tables.Signed-off-by Christopher Clark <[email protected]>---
v6 dropped Jan Acked-by due to change of implementation and commit msg v6 switched toCOMPAT_CALLand providescompat_argo_op
v6 feedback #3 Julien: addargo_opto the ARM hypercall table v2 Copyright line: add 2019
v2 feedback #3 Jan: drop "message" fromargo_message_op
v2 feedback #3 Jan: add Acked-by
v1 feedback #15 Jan: handle upper-halves of hypercall args v1 feedback #15 Jan: use unsigned where negative values impossiblexen/arch/arm/traps.c | 3 +++ xen/arch/x86/guest/hypercall_page.S | 2 +- xen/arch/x86/hvm/hypercall.c | 3 +++ xen/arch/x86/hypercall.c | 3 +++ xen/arch/x86/pv/hypercall.c | 3 +++ xen/common/Makefile | 1 + xen/common/argo.c | 38 +++++++++++++++++++++++++++++++++++++ xen/include/public/xen.h | 2 +- xen/include/xen/hypercall.h | 18 ++++++++++++++++++ 9 files changed, 71 insertions(+), 2 deletions(-) create mode 100644 xen/common/argo.cxen/arch/arm/traps.c +3
-0@@ -1397,6 +1397,9 @@ static arm_hypercall_t arm_hypercall_table[] = {HYPERCALL (platform_op,1 ),HYPERCALL_ARM (vcpu_op,3 ),HYPERCALL (vm_assist,2 ), +#ifdef CONFIG_ARGO + HYPERCALL(argo_op, 5), +#endif };#ifndef NDEBUG xen/arch/x86/guest/hypercall_page.S +1
-1@@ -59,7 +59,7 @@ DECLARE_HYPERCALL(sysctl) DECLARE_HYPERCALL(domctl) DECLARE_HYPERCALL(kexec_op) DECLARE_HYPERCALL(tmem_op) -DECLARE_HYPERCALL(xc_reserved_op) +DECLARE_HYPERCALL(argo_op) DECLARE_HYPERCALL(xenpmu_op) DECLARE_HYPERCALL(arch_0)xen/arch/x86/hvm/hypercall.c +3
-0@@ -134,6 +134,9 @@ static const hypercall_table_t hvm_hypercall_table[] = {#ifdef CONFIG_TMEM HYPERCALL(tmem_op),#endif +#ifdef CONFIG_ARGO + COMPAT_CALL(argo_op), +#endif COMPAT_CALL(platform_op),#ifdef CONFIG_PV COMPAT_CALL(mmuext_op),xen/arch/x86/hypercall.c +3
-0@@ -64,6 +64,9 @@ const hypercall_args_t hypercall_args_table[NR_hypercalls] =ARGS (domctl,1 ),ARGS (kexec_op,2 ),ARGS (tmem_op,1 ), +#ifdef CONFIG_ARGO +ARGS (argo_op,5 ), +#endifARGS (xenpmu_op,2 ), #ifdef CONFIG_HVM ARGS(hvm_op, 2),xen/arch/x86/pv/hypercall.c +3
-0@@ -77,6 +77,9 @@ const hypercall_table_t pv_hypercall_table[] = {#ifdef CONFIG_TMEM HYPERCALL(tmem_op),#endif +#ifdef CONFIG_ARGO + COMPAT_CALL(argo_op), +#endif HYPERCALL(xenpmu_op),#ifdef CONFIG_HVM HYPERCALL(hvm_op),xen/common/Makefile +1
-0@@ -1,3 +1,4 @@ +obj-$(CONFIG_ARGO) += argo.o obj-y += bitmap.o obj-y += bsearch.o obj-$(CONFIG_CORE_PARKING) += core_parking.oxen/common/argo.c new file +38
-0@@ -0,0 +1,38 @@ +/****************************************************************************** +* Argo : Hypervisor-Mediated data eXchange +* +* Derived from v4v, the version 2 of v2v. +* +* Copyright (c) 2010, Citrix Systems +* Copyright (c) 2018-2019 BAE Systems +* +* 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA +*/ + +#include <xen/errno.h> +#include <xen/guest_access.h> + +long +do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} + +#ifdef CONFIG_COMPAT +long +compat_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} +#endif xen/include/public/xen.h +1
-1@@ -118,7 +118,7 @@ DEFINE_XEN_GUEST_HANDLE(xen_ulong_t);#define __HYPERVISOR_domctl 36#define __HYPERVISOR_kexec_op 37#define __HYPERVISOR_tmem_op 38 -#define __HYPERVISOR_xc_reserved_op 39/* reserved for XenClient */ +#define __HYPERVISOR_argo_op 39#define __HYPERVISOR_xenpmu_op 40#define __HYPERVISOR_dm_op 41xen/include/xen/hypercall.h +18
-1@@ -136,6 +136,15 @@ do_tmem_op( XEN_GUEST_HANDLE_PARAM(tmem_op_t) uops);#endif +#ifdef CONFIG_ARGO +extern long do_argo_op ( +unsigned int cmd, +XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2, +unsigned long arg3, +unsigned long arg4); +#endif +extern long do_xenoprof_op (int op,XEN_GUEST_HANDLE_PARAM (void ) arg);@@ -197,6 +206,15 @@ extern int compat_multicall(XEN_GUEST_HANDLE_PARAM (multicall_entry_compat_t) call_list, uint32_t nr_calls); +#ifdef CONFIG_ARGO +extern long compat_argo_op ( +unsigned int cmd, +XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2, +unsigned long arg3, +unsigned long arg4); +#endif +#include <compat/hvm/dm_op.h> extern int --
Quoted 2 lines
2.7.4
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Hi Christopher,
Quoted 7 lines · Christopher Clark
Quoted 5 lines · Christopher Clark
Quoted 2 lines · Christopher Clark
On Thu, Jan 24, 2019 at 2:08 AM Julien Grall <[email protected]> wrote:
Quoted 8 lines
Quoted 6 lines
Quoted 3 lines
Quoted 3 lines
[...]
Sorry for noticing quite late in the process. Don't you need to add the hypercall inxen/arch/arm/traps.c?Adding this looked fine, so I've added the ARM hypercall table entry to this patch.
Julien, do you have any further feedback on the latest version of this patch?
I will not be able to look at the series this week. I defer the Arm part to Stefano.
Cheers,
Signature · 6 lines
Julien Grall _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
In reply to Christopher ClarkQuoted 10 lines · Christopher Clark
xen/common/argo.c new file +10
-0+#include <xen/errno.h> +#include <xen/guest_access.h> + +long +do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +}
xen/common/argo.c line 38
Quoted 10 lines
xen/common/argo.c new file +10
-0+ +#ifdef CONFIG_COMPAT +long +compat_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} +#endif
From an ARM perspective, it is not a good idea to use unsigned long as hypercall parameters because they are going to be of different size on arm32 and arm64. On ARM, there is no COMPAT code, and we try to keep a single stable ABI across 32bit and 64bit hypervisors (pointers size being the only exception and we deal with that using XEN_GUEST_HANDLE_PARAM).
For this reason, given that we don't need arg3 and arg4 to actually be 64bit, it would be best to use explicitly sized integers instead. I would use uint32_t or unsigned int for arg3 and arg4. That way, there are not going to be any ABI compatibility issues between arm32 and arm64 and we could run, and even migrate, 32bit guests to a 64bit hypervisor without problems.
I know that Andrew expressed concerns about using unsigned int before, but don't we just need to make sure we are properly ignoring the top 32bit of arg3 and arg4 when the hypervisor is compiled 64bit?
I am really sorry for pointing this out so late in the review cycle, but I only spotted it now.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Quoted 37 lines
Quoted 1 line
Quoted 1 line · On 05.02.19 at 01:39, <[email protected]> wrote:
Quoted 10 lines · Christopher Clark
xen/common/argo.c new file +10
-0+#include <xen/errno.h> +#include <xen/guest_access.h> + +long +do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +}xen/common/argo.c line 38
Quoted 10 lines
xen/common/argo.c new file +10
-0+ +#ifdef CONFIG_COMPAT +long +compat_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} +#endif From an ARM perspective, it is not a good idea to use unsigned long as hypercall parameters because they are going to be of different size on arm32 and arm64. On ARM, there is no COMPAT code, and we try to keep a single stable ABI across 32bit and 64bit hypervisors (pointers size being the only exception and we deal with that using
XEN_GUEST_HANDLE_PARAM).For this reason, given that we don't need arg3 and arg4 to actually be 64bit, it would be best to use explicitly sized integers instead. I would use
uint32_tor unsigned int for arg3 and arg4. That way, there are not going to be any ABI compatibility issues between arm32 and arm64 and we could run, and even migrate, 32bit guests to a 64bit hypervisor without problems.
Quoted 3 lines
I know that Andrew expressed concerns about using unsigned int before, but don't we just need to make sure we are properly ignoring the top 32bit of arg3 and arg4 when the hypervisor is compiled 64bit?
Are you saying that hypercall arguments made by a 32-bit guest on a 64-bit hypervisor do not get zero-extended before reaching the C layer (or more specifically the individual handlers, since on x86 we deal with the necessary zero-extension in C nowadays)? What about do_memory_op()'s first parameter then?
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Quoted 41 lines · Jan Beulich
Quoted 37 lines
Quoted 1 line
Quoted 1 line · On 05.02.19 at 01:39, <[email protected]> wrote:
Quoted 10 lines · Christopher Clark
xen/common/argo.c new file +10
-0+#include <xen/errno.h> +#include <xen/guest_access.h> + +long +do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +}xen/common/argo.c line 38
Quoted 10 lines
xen/common/argo.c new file +10
-0+ +#ifdef CONFIG_COMPAT +long +compat_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} +#endif From an ARM perspective, it is not a good idea to use unsigned long as hypercall parameters because they are going to be of different size on arm32 and arm64. On ARM, there is no COMPAT code, and we try to keep a single stable ABI across 32bit and 64bit hypervisors (pointers size being the only exception and we deal with that using
XEN_GUEST_HANDLE_PARAM).For this reason, given that we don't need arg3 and arg4 to actually be 64bit, it would be best to use explicitly sized integers instead. I would use
uint32_tor unsigned int for arg3 and arg4. That way, there are not going to be any ABI compatibility issues between arm32 and arm64 and we could run, and even migrate, 32bit guests to a 64bit hypervisor without problems.Quoted 3 lines
I know that Andrew expressed concerns about using unsigned int before, but don't we just need to make sure we are properly ignoring the top 32bit of arg3 and arg4 when the hypervisor is compiled 64bit?
Quoted 5 lines
Are you saying that hypercall arguments made by a 32-bit guest on a 64-bit hypervisor do not get zero-extended before reaching the C layer (or more specifically the individual handlers, since on x86 we deal with the necessary zero-extension in C nowadays)? What about do_memory_op()'s first parameter then?
If I remember right, there is no zero-extension, however, they should still be zero because they have always been zero -- nothing should change them in the VM lifetime. However, it is not great to rely on that, that is why I suggested to clear them on entry as an alternative, and also to have a single ABI between 32bit and 64bit.
FYI do_memory_op is declared as follows on the Linux side for arm32 and arm64:
int HYPERVISOR_memory_op(unsigned int cmd, void *arg);
When I went through all existing hypercalls to introduce them on arm32, I checked that we didn't actually need 64bit parameters, especially for cmd. I introduced them as int instead of long on the Linux side when possible (see include/xen/arm/hypercall.h), but I didn't attempt to modify all the existing Xen headers.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Sorry for the formatting.
Quoted 49 lines · On Tue, 5 Feb 2019, 20:04 Stefano Stabellini, <[email protected]> wrote:
Quoted 30 lines · Jan Beulich
Quoted 30 lines
Quoted 1 line
Quoted 1 line · On 05.02.19 at 01:39, <[email protected]> wrote:
Quoted 10 lines · Christopher Clark
xen/common/argo.c new file +10
-0+#include <xen/errno.h> +#include <xen/guest_access.h> + +long +do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +}xen/common/argo.c line 38
Quoted 10 lines
xen/common/argo.c new file +10
-0+ +#ifdef CONFIG_COMPAT +long +compat_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1, +XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3, +unsigned long arg4) +{ +return -ENOSYS ; +} +#endif From an ARM perspective, it is not a good idea to use unsigned long as hypercall parameters because they are going to be of different size on arm32 and arm64. On ARM, there is no COMPAT code, and we try to keep a single stable ABI across 32bit and 64bit hypervisors (pointers size being the only exception and we deal with that using
XEN_GUEST_HANDLE_PARAM).Quoted 4 lines
Quoted 4 lines
For this reason, given that we don't need arg3 and arg4 to actually be 64bit, it would be best to use explicitly sized integers instead. I would use
uint32_tor unsigned int for arg3 and arg4. That way, there are not going to be any ABI compatibility issues between arm32 andarm64
Quoted 7 lines
Quoted 6 lines
and we could run, and even migrate, 32bit guests to a 64bit hypervisor without problems.
I know that Andrew expressed concerns about using unsigned int before, but don't we just need to make sure we are properly ignoring the top 32bit of arg3 and arg4 when the hypervisor is compiled 64bit?
Quoted 5 lines
Are you saying that hypercall arguments made by a 32-bit guest on a 64-bit hypervisor do not get zero-extended before reaching the C layer (or more specifically the individual handlers, since on x86 we deal with the necessary zero-extension in C nowadays)? What about do_memory_op()'s first parameter then?
Quoted 5 lines
If I remember right, there is no zero-extension, however, they should still be zero because they have always been zero -- nothing should change them in the VM lifetime. However, it is not great to rely on that, that is why I suggested to clear them on entry as an alternative, and also to have a single ABI between 32bit and 64bit.
I can't find the wording again on the Arm Arm (the pdf reader on the phone is not great).
But I am afraid this is not correct. Upper 32-bit of the register will be zeroed when writing a 32-bit value. So we never rely on the register to be zeroed on boot.
Quoted 11 lines
FYI
do_memory_opis declared as follows on the Linux side for arm32 and arm64:int HYPERVISOR_memory_op(unsigned int cmd, void *arg);When I went through all existing hypercalls to introduce them on arm32, I checked that we didn't actually need 64bit parameters, especially for cmd. I introduced them as int instead of long on the Linux side when possible (see
include/xen/arm/hypercall.h), but I didn't attempt to modify all the existing Xen headers.
I don't understand your concern with unsigned long. We use them in __DEFINE_XEN_GUEST_HANDLE that are in turn to describe guest pointer.
The problem with explicitly sized (i.e 32-bit) is you ignore the top 32-bit. This means you can't check the upper bits are always 0 and would prevent extension.
Cheers,
Quoted 4 lines
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Quoted 58 lines · Julien Grall
Sorry for the formatting.
On Tue, 5 Feb 2019, 20:04 Stefano Stabellini, <[email protected]> wrote:
On Tue, 5 Feb 2019, Jan Beulich wrote:
> >>> On 05.02.19 at 01:39, <[email protected]> wrote:
> > On Wed, 30 Jan 2019, Christopher Clark wrote:
> >> +#include <xen/errno.h>
> >> +#include <xen/guest_access.h>
> >> +
> >> +long
> >> +do_argo_op(unsigned int cmd,XEN_GUEST_HANDLE_PARAM(void)arg1, > >> + XEN_GUEST_HANDLE_PARAM(void) arg2, unsigned long arg3, > >> + unsigned long arg4)
> >> +{
> >> + return-ENOSYS;
> >> +}
> >> +
> >> +#ifdefCONFIG_COMPAT
> >> +long
> >> +compat_argo_op(unsigned int cmd,XEN_GUEST_HANDLE_PARAM(void)arg1, > >> + XEN_GUEST_HANDLE_PARAM(void) arg2, unsigned long arg3, > >> + unsigned long arg4)
> >> +{
> >> + return-ENOSYS;
> >> +}
> >> +#endif
> >
> > From an ARM perspective, it is not a good idea to use unsigned long as > > hypercall parameters because they are going to be of different size on > > arm32 and arm64. On ARM, there is no COMPAT code, and we try to keep a > > single stable ABI across 32bit and 64bit hypervisors (pointers size > > being the only exception and we deal with that using
> >XEN_GUEST_HANDLE_PARAM).
> >
> > For this reason, given that we don't need arg3 and arg4 to actually be > > 64bit, it would be best to use explicitly sized integers instead. I > > would useuint32_tor unsigned int for arg3 and arg4. That way, there > > are not going to be any ABI compatibility issues between arm32 and arm64 > > and we could run, and even migrate, 32bit guests to a 64bit hypervisor > > without problems.
> >
> > I know that Andrew expressed concerns about using unsigned int before, > > but don't we just need to make sure we are properly ignoring the top > > 32bit of arg3 and arg4 when the hypervisor is compiled 64bit? >
> Are you saying that hypercall arguments made by a 32-bit guest on a > 64-bit hypervisor do not get zero-extended before reaching the C layer > (or more specifically the individual handlers, since on x86 we deal with > the necessary zero-extension in C nowadays)? What about
> do_memory_op()'s first parameter then?If I remember right, there is no zero-extension, however, they should still be zero because they have always been zero -- nothing should change them in the VM lifetime. However, it is not great to rely on that, that is why I suggested to clear them on entry as an alternative, and also to have a single ABI between 32bit and 64bit.
Quoted 4 lines
I can't find the wording again on the Arm Arm (the pdf reader on the phone is not great).
But I am afraid this is not correct. Upper 32-bit of the register will be zeroed when writing a 32-bit value. So we never rely on the register to be zeroed on boot.
Hi Julien,
Thank you for checking your emails. I found the reference in the ARM ARM, although it took me several minutes!
"The upper 32 bits of the destination register are set to zero."
from C6.1.1 (ID092916).
Quoted 12 lines
FYI
do_memory_opis declared as follows on the Linux side for arm32 and arm64:int
HYPERVISOR_memory_op(unsigned int cmd, void *arg);When I went through all existing hypercalls to introduce them on arm32, I checked that we didn't actually need 64bit parameters, especially for cmd. I introduced them as int instead of long on the Linux side when possible (see
include/xen/arm/hypercall.h), but I didn't attempt to modify all the existing Xen headers.
Quoted 2 lines
I don't understand your concern with unsigned long. We use them in
__DEFINE_XEN_GUEST_HANDLEthat are in turn to describe guest pointer.
__DEFINE_XEN_GUEST_HANDLE is for pointers inside memory structs, and we defined it as:
* XEN_GUEST_HANDLE represents a guest pointer, when passed as a field
* in a struct in memory. On ARM is always 8 bytes sizes and 8 bytes
* aligned.
You probably meant XEN_GUEST_HANDLE_PARAM which is the one for pointers when passed as hypercall parameters, that is defined as:
* XEN_GUEST_HANDLE_PARAM represents a guest pointer, when passed as an
* hypercall argument. It is 4 bytes on aarch32 and 8 bytes on aarch64.
Yes, pointers as hypercalls parameters are the exception to the single-ABI rule and we introduced XEN_GUEST_HANDLE_PARAM purposely to handle them. However, I am not sure we took into account zero-extension when we discussed hypercalls parameters for arm back in the day when I wrote include/xen/arm/hypercall.h.
Quoted 2 lines
The problem with explicitly sized (i.e 32-bit) is you ignore the top 32-bit. This means you can't check the upper bits are always 0 and would prevent extension.
That is true. I implicitly assumed that our desire for a common 32-bit/64-bit ABI would not apply just to structs in memory (where we always define unsigned long and pointers as 64-bit) but also seamlessly apply to hypercalls parameters (except for pointers as per the above).
There are still reasons for choosing unsigned int for cases like this where unsigned long is not actually necessary, but not a strong as I previously thought. For example, it could be natural to introduce a value for a cmd or a flag parameter not available to 32-bit guests (i.e. 0xff00000000000000) by mistake, although I admit that the related Xen code should throw a warning when compiled for arm32.
In conclusion, if you and other maintainers prefer unsigned long I'll drop my reservation.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Hi Stefano,
Quoted 10 lines · Stefano Stabellini
Quoted 7 lines · Julien Grall
Sorry for the formatting.
On Tue, 5 Feb 2019, 20:04 Stefano Stabellini, <[email protected]> wrote:
On Tue, 5 Feb 2019, Jan Beulich wrote:But I am afraid this is not correct. Upper 32-bit of the register will be zeroed when writing a 32-bit value. So we never rely on the register to be zeroed on boot.
Hi Julien,
Hi Stefano,
Quoted 7 lines
Thank you for checking your emails. I found the reference in the ARM ARM, although it took me several minutes!
"The upper 32 bits of the destination register are set to zero."
from C6.1.1 (ID092916).
Actually, you were right and I was wrong. This paragraph only talks about 32-bit access from AArch64. I found a paragraph on the Arm Arm (D1.20.2 in DDI 0487D.a) stating that the upper 32-bits can either "be zeroed, or hold the value that the same architectural register held before any AArch32".
So we do rely the upper bits are zeroed when the vCPU is created. The registers are set by arch_set_info_guest via a context. The context can be set by either virtual PSCI call CPU_ON or via DOMCTL setvcpucontext (so the tools).
We fully control PSCI call CPU_ON and zero the registers. So no question here.
For the DOMCTL, per XSA-77, we trust how operation will be used. The toolstack (vcpu_arm32 libxc/xc_dom_arm.c) will zero the context before. So I think we should be safe here.
However, I think we should add some sanity check in arch_set_info_guest for our peace of mind. For guest entry/exit, rather than zero the upper 32-bits I would also add sanity check in enter_hypervisor_head and leave_hypervisor_tail but only in debug build. Any opinions?
Quoted 21 lines
Quoted 12 lines
FYI
do_memory_opis declared as follows on the Linux side for arm32 and arm64:int
HYPERVISOR_memory_op(unsigned int cmd, void *arg);When I went through all existing hypercalls to introduce them on arm32, I checked that we didn't actually need 64bit parameters, especially for cmd. I introduced them as int instead of long on the Linux side when possible (see
include/xen/arm/hypercall.h), but I didn't attempt to modify all the existing Xen headers.Quoted 2 lines
I don't understand your concern with unsigned long. We use them in
__DEFINE_XEN_GUEST_HANDLEthat are in turn to describe guest pointer.
__DEFINE_XEN_GUEST_HANDLEis for pointers inside memory structs, and we defined it as:
*XEN_GUEST_HANDLErepresents a guest pointer, when passed as a field
* in a struct in memory. On ARM is always 8 bytes sizes and 8 bytes
* aligned.
Quoted 4 lines
You probably meant
XEN_GUEST_HANDLE_PARAMwhich is the one for pointers when passed as hypercall parameters, that is defined as:
*XEN_GUEST_HANDLE_PARAMrepresents a guest pointer, when passed as an
* hypercall argument. It is 4 bytes on aarch32 and 8 bytes on aarch64.
Yes, I meant XEN_GUEST_HANDLE_PARAM. I should have waited to have my laptop in hand rather looking on my phone :).
Quoted 6 lines
Yes, pointers as hypercalls parameters are the exception to the single-ABI rule and we introduced
XEN_GUEST_HANDLE_PARAMpurposely to handle them. However, I am not sure we took into account zero-extension when we discussed hypercalls parameters for arm back in the day when I wroteinclude/xen/arm/hypercall.h.
I am not sure to follow your thoughts about taking into account zero-extension in the Linux header. Could you expand it?
In the official public headers, I can't find anything telling you the size of each arguments and the number of arguments. Instead you have to look at Xen code to know the exact number of arguments and the size. Did I miss anything?
Regarding Argo, there seem to have some kind of documentation per-hypercalls although some does not specify the size. But I can't find anything telling you the command number is in arg0. The mapping to from argN the hypercalls arguments is also not that clear.
But I guess this kind of improvement can be done afterwards.
Quoted 7 lines
Quoted 2 lines
The problem with explicitly sized (i.e 32-bit) is you ignore the top 32-bit. This means you can't check the upper bits are always 0 and would prevent extension.
That is true. I implicitly assumed that our desire for a common 32-bit/64-bit ABI would not apply just to structs in memory (where we always define unsigned long and pointers as 64-bit) but also seamlessly apply to hypercalls parameters (except for pointers as per the above).
There are no documentation in the public interface about the size of each arguments. When looking at traps.c, we assume that hypercalls arguments are the size of a register:
typedef register_t (*arm_hypercall_fn_t)(register_t, register_t, register_t, register_t, register_t).
So for 64-bit Xen, the hypercalls arguments will be 64-bit.
If we want to impose 32-bit value, then we need to update the callback, add sanity check (?) and probably document it.
Quoted 7 lines
There are still reasons for choosing unsigned int for cases like this where unsigned long is not actually necessary, but not a strong as I previously thought. For example, it could be natural to introduce a value for a cmd or a flag parameter not available to 32-bit guests (i.e.
0xff00000000000000) by mistake, although I admit that the related Xen code should throw a warning when compiled for arm32.
I think the compiler should catch it. However, I think allowing up to 64-bit arguments is nice to have if we want to introduce extension for 64-bit only guest (i.e larger buffer...).
Quoted 3 lines
In conclusion, if you and other maintainers prefer unsigned long I'll drop my reservation.
The summary of my e-mail is:
- we need to add sanity check (or zero) upper-bits for 32-bit guest
- unsigned long should be fine for 64-bit only features
- we need to document the behavior of each hypercall and provide
guidelines for new one.
None of this is specific to Argo and I would be happy to defer this as a follow-up series.
Cheers,
Signature · 6 lines
Julien Grall _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Quoted 17 lines · Julien Grall
Hi Stefano,
Quoted 14 lines · Stefano Stabellini
Quoted 9 lines · Julien Grall
Sorry for the formatting.
On Tue, 5 Feb 2019, 20:04 Stefano Stabellini, <[email protected]> wrote:
On Tue, 5 Feb 2019, Jan Beulich wrote:But I am afraid this is not correct. Upper 32-bit of the register will be zeroed when writing a 32-bit value. So we never rely on
the register to be zeroed on boot.Thank you for checking your emails. I found the reference in the ARM ARM, although it took me several minutes!
Quoted 8 lines
Quoted 3 lines
"The upper 32 bits of the destination register are set to zero."
from C6.1.1 (ID092916).
Actually, you were right and I was wrong. This paragraph only talks about 32-bit access from AArch64. I found a paragraph on the Arm Arm (D1.20.2 in DDI 0487D.a) stating that the upper 32-bits can either "be zeroed, or hold the value that the same architectural register held before any AArch32".
Understanding the ARM ARM is really difficult, I am glad we clarified this (and that my memory didn't completely fail me yet).
Quoted 7 lines
So we do rely the upper bits are zeroed when the vCPU is created. The registers are set by
arch_set_info_guestvia a context. The context can be set by either virtual PSCI callCPU_ONor via DOMCTL setvcpucontext (so the tools).We fully control PSCI call
CPU_ONand zero the registers. So no question here.
Quoted 8 lines
For the DOMCTL, per XSA-77, we trust how operation will be used. The toolstack (
vcpu_arm32libxc/xc_dom_arm.c) will zero the context before. So I think we should be safe here.However, I think we should add some sanity check in
arch_set_info_guestfor our peace of mind. For guest entry/exit, rather than zero the upper 32-bits I would also add sanity check inenter_hypervisor_headandleave_hypervisor_tailbut only in debug build. Any opinions?
Definitely we should have sanity checks.
Quoted 41 lines
Quoted 26 lines
Quoted 16 lines
FYI
do_memory_opis declared as follows on the Linux side for arm32
andarm64: int HYPERVISOR_memory_op(unsigned int cmd, void *arg); When I went through all existing hypercalls to introduce them onarm32,
I checked that we didn't actually need 64bit parameters, especially
for
cmd. I introduced them as int instead of long on the Linux side
when
possible (seeinclude/xen/arm/hypercall.h), but I didn't attempt to modify all the existing Xen headers.Quoted 3 lines
I don't understand your concern with unsigned long. We use them in
__DEFINE_XEN_GUEST_HANDLEthat are in turn to describe guest pointer.
__DEFINE_XEN_GUEST_HANDLEis for pointers inside memory structs, and we defined it as:
*XEN_GUEST_HANDLErepresents a guest pointer, when passed as a field
* in a struct in memory. On ARM is always 8 bytes sizes and 8 bytes
* aligned.Quoted 4 lines
You probably meant
XEN_GUEST_HANDLE_PARAMwhich is the one for pointers when passed as hypercall parameters, that is defined as:
*XEN_GUEST_HANDLE_PARAMrepresents a guest pointer, when passed as an
* hypercall argument. It is 4 bytes on aarch32 and 8 bytes on aarch64.Yes, I meant
XEN_GUEST_HANDLE_PARAM. I should have waited to have my laptop in hand rather looking on my phone :).Quoted 6 lines
Yes, pointers as hypercalls parameters are the exception to the single-ABI rule and we introduced
XEN_GUEST_HANDLE_PARAMpurposely to handle them. However, I am not sure we took into account zero-extension when we discussed hypercalls parameters for arm back in the day when I wroteinclude/xen/arm/hypercall.h.
Quoted 6 lines
I am not sure to follow your thoughts about taking into account zero-extension in the Linux header. Could you expand it?
In the official public headers, I can't find anything telling you the size of each arguments and the number of arguments. Instead you have to look at Xen code to know the exact number of arguments and the size. Did I miss anything?
No, it is like you wrote. I think I should have pushed the discussion further and added more information to the Xen public headers back in those days.
Quoted 20 lines
Regarding Argo, there seem to have some kind of documentation per-hypercalls although some does not specify the size. But I can't find anything telling you the command number is in arg0. The mapping to from argN the hypercalls arguments is also not that clear.
But I guess this kind of improvement can be done afterwards.
Quoted 8 lines
Quoted 3 lines
The problem with explicitly sized (i.e 32-bit) is you ignore the top 32-bit. This means you can't check the upper bits are always 0 and would prevent extension.
That is true. I implicitly assumed that our desire for a common 32-bit/64-bit ABI would not apply just to structs in memory (where we always define unsigned long and pointers as 64-bit) but also seamlessly apply to hypercalls parameters (except for pointers as per the above).
There are no documentation in the public interface about the size of each arguments. When looking at
traps.c, we assume that hypercalls arguments are the size of a register:
Quoted 7 lines
typedef
register_t(*arm_hypercall_fn_t)(register_t,register_t,register_t,register_t,register_t).So for 64-bit Xen, the hypercalls arguments will be 64-bit.
If we want to impose 32-bit value, then we need to update the callback, add sanity check (?) and probably document it.
Good point.
Quoted 10 lines
Quoted 6 lines
There are still reasons for choosing unsigned int for cases like this where unsigned long is not actually necessary, but not a strong as I previously thought. For example, it could be natural to introduce a value for a cmd or a flag parameter not available to 32-bit guests (i.e.
0xff00000000000000) by mistake, although I admit that the related Xen code should throw a warning when compiled for arm32.I think the compiler should catch it. However, I think allowing up to 64-bit arguments is nice to have if we want to introduce extension for 64-bit only guest (i.e larger buffer...).
That's true. In this case, Christopher was telling me the arguments could easily be 32bit only.
Quoted 6 lines
Quoted 2 lines
In conclusion, if you and other maintainers prefer unsigned long I'll drop my reservation.
The summary of my e-mail is:
- we need to add sanity check (or zero) upper-bits for 32-bit guest
- unsigned long should be fine for 64-bit only features
I take that you mean that unsigned long should be fine, and would also allow us to introduce 64-bit only features in the future, right? You are *not* saying that unsigned long should only be used with 64-bit guests/hypervisor?
Quoted 5 lines
- we need to document the behavior of each hypercall and provide
guidelines for new one.None of this is specific to Argo and I would be happy to defer this as a follow-up series.
Assuming my understanding is right, I agree with you.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 02/15] argo: introduce the argo_op hypercall boilerplate
Hi Stefano,
Quoted 7 lines · Stefano Stabellini
Quoted 4 lines · Julien Grall
However, I think we should add some sanity check in
arch_set_info_guestfor our peace of mind. For guest entry/exit, rather than zero the upper 32-bits I would also add sanity check inenter_hypervisor_headandleave_hypervisor_tailbut only in debug build. Any opinions?Definitely we should have sanity checks.
I will add this one and ...
Quoted 7 lines
Quoted 3 lines
In the official public headers, I can't find anything telling you the size of each arguments and the number of arguments. Instead you have to look at Xen code to know the exact number of arguments and the size. Did I miss anything?
No, it is like you wrote. I think I should have pushed the discussion further and added more information to the Xen public headers back in those days.
... and this one in my list of things to improve.
Quoted 9 lines
Quoted 6 lines
Quoted 2 lines
In conclusion, if you and other maintainers prefer unsigned long I'll drop my reservation.
The summary of my e-mail is:
- we need to add sanity check (or zero) upper-bits for 32-bit guest
- unsigned long should be fine for 64-bit only featuresI take that you mean that unsigned long should be fine, and would also allow us to introduce 64-bit only features in the future, right?
that's correct.
You are
Quoted 2 lines
*not* saying that unsigned long should only be used with 64-bit guests/hypervisor?
unsigned long == uint32_t on 32-bit. So it would be silly of me to suggest that :).
Quoted 8 lines
Quoted 6 lines
- we need to document the behavior of each hypercall and provide
guidelines for new one.None of this is specific to Argo and I would be happy to defer this as a follow-up series.
Assuming my understanding is right, I agree with you.
You understood right.
Cheers,
Signature · 6 lines
Julien Grall _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 3/15 argo: define argo_dprintk for subsystem debugging
In reply to Christopher ClarkA convenience for working on development of the argo subsystem:
setting a #define variable enables additional debug messages.
---
v3 added Roger's Reviewed-by
v3 added Jan's Ack
v2 #03 feedback, Jan: fix ifdef/define confusion error v1 #04 feedback, Jan: fix dprintk implementation
xen/common/argo.c | 9 +++++++++ 1 file changed, 9 insertions(+)
xen/common/argo.c +9 -0
@@ -19,6 +19,15 @@#include <xen/errno.h> #include <xen/guest_access.h> +/* Change this to #define ARGO_DEBUG here to enable more debug messages */ +#undef ARGO_DEBUG + +#ifdef ARGO_DEBUG +#define argo_dprintk(format, args...) printk("argo: " format, ## args ) +#else +#define argo_dprintk (format, ... ) ((void)0) +#endif + long do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, XEN_GUEST_HANDLE_PARAM(void) arg2, unsigned long arg3,
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 4/15 argo: init, destroy and soft-reset, with enable command line opt
In reply to Christopher ClarkInitialises basic data structures and performs teardown of argo state for domain shutdown.
Inclusion of the Argo implementation is dependent on CONFIG_ARGO.
Introduces a new Xen command line parameter 'argo': bool to enable/disable the argo hypercall. Defaults to disabled.
New headers:
public/argo.h: with definions of addresses and ring structure, including indexes for atomic update for communication between domain and hypervisor.
xen/argo.h: to expose the hooks for integration into domain lifecycle:
argo_init: per-domain init of argo data structures for domain_create.
argo_destroy: teardown for domain_destroy and the error exit
path of domain_create.
argo_soft_reset: reset of domain state for domain_soft_reset.
Adds a new field to struct domain: struct argo_domain *argo;
In accordance with recent work on _domain_destroy, argo_destroy is idempotent. It will tear down: all rings registered by this domain, all rings where this domain is the single sender (ie. specified partner, non-wildcard rings), and all pending notifications where this domain is awaiting signal about available space in the rings of other domains.
A count will be maintained of the number of rings that a domain has registered in order to limit it below the fixed maximum limit defined here.
Macros are defined to verify the internal locking state within the argo implementation. The macros are ASSERTed on entry to functions to validate and document the required lock state prior to calling.
The hash function for the hashtables that hold ring state is derived from the string hashing function djb2 (http://www.cse.yorku.ca/~oz/hash.html)
by Daniel J. Bernstein. Basic testing with a limited number of domains and ports has shown reasonable distribution for the table size.
The software license on the public header is the BSD license, standard procedure for the public Xen headers. The public header was originally posted under a GPL license at: [1]:
https://lists.xenproject.org/archives/html/xen-devel/2013-05/msg02710.html
The following ACK by Lars Kurth is to confirm that only people being employees of Citrix contributed to the header files in the series posted at
[1] and that thus the copyright of the files in question is fully owned by Citrix. The ACK also confirms that Citrix is happy for the header files to be published under a BSD license in this series (which is based on [1]).
---
v6 #09 Jan: introduce compat ABI
v6 #04 Jan: xlat.lst: move argo struct entries to alphabetical position v6 #04 Roger: use list_for_each_entry{_safe} for looping
v5 #04 Roger: tweak command line doc: remove statement about top level bool
v5: add compat validation macros to primary source file: common/argo.c
v5: dropped external file for compat macros: common/compat/argo.c
v4: removed FIXME for removing argo_destroy from domain_kill
v4 Jan: amend the command line doc text referring to build configuration v4 : use standard data structures as per common code
v4 Jan: replace hash_index with djb2-derived hash algorithm
v4 Andrew: switch argo command line option to list argo=<bool>
v4 #04 Roger: drop unneeded init of ring_count in argo_domain_init v4 #04 Roger: replace if (ring_info->mfns) with ASSERTs in ring_unmap v4 #04 Roger: rewrite the locking verification macros
v4 #04 Roger: make L1 lock description comment clearer about R(L1) and W(L1) v4 Andrew: fix split of dprintk in ring_map_info across v4 commits v3 #04 Andrew: use xzalloc for struct argo_domain in argo_init
v3 #04 Andrew: reference CONFIG_ARGO in the command line documentation v3 #07 Jan: rename ring_find_info to find_ring_info
v3 #04 Andrew: don't truncate args do_argo_op printk
v3 #07 Jan: fix numeric entries in printk format strings
v3 #10 Roger: move find functions to top of file and drop prototypes v3 #04 Jan: meld compat check for hypercall arg types
v3 #04 Roger/Jan: make lock names clearer and assert their state
v3 #04 Jan: port -> aport with type; distinguish argo port from evtchn v3 #04 Jan: reorder call to argo_init_domain in argo_init
v3 #04 Jan: ring_remove_mfns: zero count before freeing arrays
v3 #04 Jason/Roger: soft_reset: can assume reinit is ok if d->argo set v3 #04 Roger: remove unused and confusing d->argo_lock
v3 #04 Roger: add simple inlines in xen/argo.h, drop ifdef CONFIG_ARGO v3 #04 Roger: simpler return -EOPNOTSUPP in do_argo_op
v3 #04 Roger: add const to domain arg to ring_remove_info
v3 #04 Roger: use XFREE
v3 #04 Roger: newline fix in wildcard_pending_list_remove
v3 #04 Roger: mfn_mapping: void* instead of uint8_t*
v3 #04 Roger: drop npages struct member in argo_ring_info; use len v3 #04 Roger/Jan: drop many fixed width types in internal structs v3 #04 Jason/Jan: drop pad and fixed width type in pending_ent struct v3 #04 Eric: moved ring_find_info from register op into this commit v3 moved hash_index function, nospec include from register op to this commit v3 moved XEN_ARGO_DOMID_ANY defn from register op into this commit v3 added #include <xen/sched.h> to <xen/argo.h> for domain struct defn v3 feedback #04 Roger: reorder #includes to alphabetical order
v3 Added Ross's Reviewed-by.
v2 rewrite locking explanation comment
v2 header copyright line now includes 2019
v2 self: use ring_info backpointer in pending_ent to maintain npending v2 self: rename all_rings_remove_info to domain_rings_remove_all
v2 feedback Jan: drop cookie, implement teardown
v2 self: add npending to track number of pending entries per ring v2 self: amend comment on locking; drop section comments
v2 cookie_eq: test low bits first and use likely on high bits
v2 self: OVERHAUL
v2 self: s/argo_pending_ent/pending_ent/g
v2 self: drop pending_remove_ent, inline at single call site
v1 feedback Roger, Jan: drop argo prefix on static functions
v2 #4 Lars: add Acked-by and details to commit message.
v2 feedback #9 Jan: document argo boot opt in xen-command-line.markdown v2 bugfix: xsm use in soft-reset prior to introduction
v2 feedback #9 Jan: drop 'message' from do_argo_message_op
v1 #5 feedback Paul: init/destroy unsigned, brackets and whitespace fixes v1 #5 feedback Paul: Use mfn_eq for comparing mfns.
v1 #5 feedback Paul: init/destroy : use currd
v1 #6 (#5) feedback Jan: init/destroy: s/ENOSYS/EOPNOTSUPP/
v1 #6 feedback Paul: Folded patch 6 into patch 5.
v1 #6 feedback Jan: drop opt_argo_enabled initializer
v1 $6 feedback Jan: s/ENOSYS/EOPNOTSUPP/g and drop useless dprintk v1. #5 feedback Paul: change the license on public header to BSD
- ack from Lars at Citrix.
v1. self, Jan: drop unnecessary xen include from sched.h
v1. self, Jan: drop inclusion of public argo.h in private one
v1. self, Jan: add include of public argo.h to argo.c
v1. self, Jan: drop fwd decl of argo_domain in priv header
v1. Paul/self/Jan: add data structures to xlat.lst and compat/argo.h to Makefile v1. self: removed allocation of event channel since switching to VIRQ v1. self: drop types.h include from private argo.h
v1: reorder public argo include position
v1: #13 feedback Jan: public namespace: prefix with xen
v1: self: rename pending ent "id" to "domain_id"
v1: self: add domain_cookie to ent struct
v1. #15 feedback Jan: make cmd unsigned
v1. #15 feedback Jan: make i loop variable unsigned
v1: self: adjust dprintks in init, destroy
v1: #18 feedback Jan: meld max ring count limit
v1: self: use type not struct in public defn, affects compat gen header
v1: feedback #15 Jan: handle upper-halves of hypercall args
v1: add comment explaining the 'magic' field
v1: self + Jan feedback: implement soft reset
v1: feedback #13 Roger: use ASSERT_UNREACHABLE
docs/misc/xen-command-line.pandoc | 13 + xen/common/argo.c | 630 +++++++++++++++++++++++++++++++++++++- xen/common/domain.c | 9 + xen/include/Makefile | 1 + xen/include/public/argo.h | 64 ++++ xen/include/xen/argo.h | 44 +++ xen/include/xen/sched.h | 5 + xen/include/xlat.lst | 2 + 8 files changed, 766 insertions(+), 2 deletions(-) create mode 100644 xen/include/public/argo.h create mode 100644 xen/include/xen/argo.h
docs/misc/xen-command-line.pandoc +13 -0
@@ -182,6 +182,19 @@ Permit Xen to use "Always Running APIC Timer" support on compatible hardware in combination with cpuidle. This option is only expected to be useful for developers wishing Xen to fall back to older timing methods on newer hardware. +### argo + = List of [ <bool> ] + +Controls for the Argo hypervisor-mediated interdomain communication service. + +The functionality that this option controls is only available when Xen has been +compiled with the build setting for Argo enabled in the build configuration. + +Argo is a interdomain communication mechanism, where Xen acts as the central +point of authority. Guests may register memory rings to recieve messages, +query the status of other domains, and send messages by hypercall, all subject +to appropriate auditing by Xen. Argo is disabled by default. + ### asid (x86) > `= <boolean>`
xen/common/argo.c +628 -2
654 lines not on this page · raw message
xen/common/domain.c +9 -0
@@ -32,6 +32,7 @@#include <xen/grant_table.h> #include <xen/xenoprof.h> #include <xen/irq.h> +#include <xen/argo.h> #include <asm/debugger.h> #include <asm/p2m.h> #include <asm/processor.h>
@@ -277,6 +278,8 @@ static void _domain_destroy(struct domain *d)xfree (d->pbuf); +argo_destroy (d); +rangeset_domain_destroy (d);free_cpumask_var (d->dirty_cpumask);
@@ -445,6 +448,9 @@ struct domain *domain_create(domid_t domid, goto fail; init_status |= INIT_gnttab; +if ( (err =argo_init (d)) !=0 ) + goto fail; + err = -ENOMEM ; d->pbuf =xzalloc_array (char,DOMAIN_PBUF_SIZE );
@@ -717,6 +723,7 @@ int domain_kill(struct domain *d)if ( d->is_dying != DOMDYING_alive )return domain_kill (d); d->is_dying = DOMDYING_dying; +argo_destroy (d);evtchn_destroy (d);gnttab_release_mappings (d);tmem_destroy (d->tmem_client);
@@ -1175,6 +1182,8 @@ int domain_soft_reset(struct domain *d)grant_table_warn_active_grants (d); +argo_soft_reset (d); + for_each_vcpu ( d, v ) {set_xen_guest_handle (runstate_guest (v),NULL );
xen/include/Makefile +1 -0
@@ -5,6 +5,7 @@ ifneq ($(CONFIG_COMPAT),) compat-arch-$(CONFIG_X86) := x86_32 headers-y := \ + compat/argo.h \ compat/callback.h \ compat/elfnote.h \ compat/event_channel.h \
xen/include/public/argo.h new file +64 -0
64 lines not on this page · raw message
xen/include/xen/argo.h new file +44 -0
44 lines not on this page · raw message
xen/include/xen/sched.h +5 -0
@@ -490,6 +490,11 @@ struct domain unsigned int guest_request_enabled : 1; unsigned int guest_request_sync : 1; } monitor; + +#ifdef CONFIG_ARGO +/* Argo interdomain communication support */ +struct argo_domain *argo; +#endif };/* Protect updates/reads (resp.) of domain_list and domain_hash. */
xen/include/xlat.lst +2 -0
@@ -30,6 +30,8 @@ ? mc_notifydomain arch-x86/xen-mca.h ! mc_physcpuinfo arch-x86/xen-mca.h ? page_offline_action arch-x86/xen-mca.h +? argo_addr argo.h +? argo_ring argo.h ? evtchn_alloc_unbound event_channel.h ? evtchn_bind_interdomain event_channel.h ? evtchn_bind_ipi event_channel.h
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 04/15] argo: init, destroy and soft-reset, with enable command line opt
Quoted 49 lines · Christopher Clark
Initialises basic data structures and performs teardown of argo state for domain shutdown.
Inclusion of the Argo implementation is dependent on
CONFIG_ARGO.Introduces a new Xen command line parameter 'argo': bool to enable/disable the argo hypercall. Defaults to disabled.
New headers:
public/argo.h: with definions of addresses and ring structure, including indexes for atomic update for communication between domain and hypervisor.
xen/argo.h: to expose the hooks for integration into domain lifecycle:
argo_init: per-domain init of argo data structures fordomain_create.
argo_destroy: teardown fordomain_destroyand the error exit
path ofdomain_create.
argo_soft_reset: reset of domain state fordomain_soft_reset.Adds a new field to struct domain: struct
argo_domain*argo;In accordance with recent work on
_domain_destroy,argo_destroyis idempotent. It will tear down: all rings registered by this domain, all rings where this domain is the single sender (ie. specified partner, non-wildcard rings), and all pending notifications where this domain is awaiting signal about available space in the rings of other domains.A count will be maintained of the number of rings that a domain has registered in order to limit it below the fixed maximum limit defined here.
Macros are defined to verify the internal locking state within the argo implementation. The macros are ASSERTed on entry to functions to validate and document the required lock state prior to calling.
The hash function for the hashtables that hold ring state is derived from the string hashing function djb2 (http://www.cse.yorku.ca/~oz/hash.html)
by Daniel J. Bernstein. Basic testing with a limited number of domains and ports has shown reasonable distribution for the table size.The software license on the public header is the BSD license, standard procedure for the public Xen headers. The public header was originally posted under a GPL license at: [1]:
https://lists.xenproject.org/archives/html/xen-devel/2013-05/msg02710.htmlThe following ACK by Lars Kurth is to confirm that only people being employees of Citrix contributed to the header files in the series posted at
[1] and that thus the copyright of the files in question is fully owned by Citrix. The ACK also confirms that Citrix is happy for the header files to be published under a BSD license in this series (which is based on [1]).
Quoted 4 lines
Signed-off-by Christopher Clark <[email protected]>Acked-by Lars Kurth <[email protected]>Reviewed-by Ross Philipson <[email protected]>Tested-by Chris Patterson <[email protected]>
I have some comments below that could be fixed upon commit if the committer agrees.
xen/common/argo.c line 324
Quoted 12 lines
xen/common/argo.c +12
-0+static struct argo_ring_info * +find_ring_info (const struct domain *d,const struct argo_ring_id *id) +{ +struct argo_ring_info *ring_info; +const struct list_head *bucket; + +ASSERT (LOCKING_Read_rings_L2 (d)); + +/* List is not modified here. Search and return the match if found. */ + bucket = &d->argo->ring_hash[hash_index (id)]; + +list_for_each_entry (ring_info, bucket, node)
I'm not sure what's the policy regarding list_ macros, should spaces be added between the parentheses?
list_for_each_entry ( ring_info, bucket, node )
I don't have a strong opinion either way.
[...]
xen/common/argo.c line 396
Quoted 9 lines
xen/common/argo.c +9
-0+static void +pending_remove_all (const struct domain *d,struct argo_ring_info *ring_info) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* Delete all pending notifications from this ring's list. */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node)
list_first_entry_or_null and you can get rid of next.
Quoted 9 lines
xen/common/argo.c +9
-0+ { +/* For wildcard rings, remove each from their wildcard list too. */ +if ( ring_info->id.partner_id ==XEN_ARGO_DOMID_ANY ) +wildcard_pending_list_remove (ent->domain_id, ent); +list_del (&ent->node); +xfree (ent); + } + ring_info->npending =0 ; +}
xen/common/argo.c line 415
Quoted 10 lines
xen/common/argo.c +10
-0+ +static void +wildcard_rings_pending_remove (struct domain *d) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_Write_L1); + +/* Delete all pending signals to the domain about wildcard rings. */ +list_for_each_entry_safe (ent, next, &d->argo->wildcard_pend_list, node)
list_first_entry_or_null and you can get rid of next.
Quoted 55 lines
xen/common/argo.c +55
-0+ { +/* +* The ent->node deleted here, and the npending value decreased, +* belong to the ring_info of another domain, which is why this +* function requires holding W(L1): +* it implies the L3 lock that protects that ring_info struct. +*/ + ent->ring_info->npending--; +list_del (&ent->node); +list_del (&ent->wildcard_node); +xfree (ent); + } +} + +static void +ring_remove_mfns (const struct domain *d,struct argo_ring_info *ring_info) +{ +unsigned int i; + +ASSERT (LOCKING_Write_rings_L2 (d)); + +if ( !ring_info->mfns ) +return ; + +if ( !ring_info->mfn_mapping ) + { +ASSERT_UNREACHABLE (); +return ; + } + +ring_unmap (d, ring_info); + +for ( i =0 ; i < ring_info->nmfns; i++ ) +if ( !mfn_eq (ring_info->mfns[i],INVALID_MFN ) ) +put_page_and_type (mfn_to_page (ring_info->mfns[i])); + + ring_info->nmfns =0 ; +XFREE (ring_info->mfns); +XFREE (ring_info->mfn_mapping); +} + +static void +ring_remove_info (const struct domain *d,struct argo_ring_info *ring_info) +{ +ASSERT (LOCKING_Write_rings_L2 (d)); + +pending_remove_all (d, ring_info); +list_del (&ring_info->node); +ring_remove_mfns (d, ring_info); +xfree (ring_info); +} + +static void +domain_rings_remove_all (struct domain *d) +{
xen/common/argo.c line 480
Quoted 10 lines
xen/common/argo.c +10
-0+unsigned int i; + +ASSERT (LOCKING_Write_rings_L2 (d)); + +for ( i =0 ; i <ARGO_HASHTABLE_SIZE ; ++i ) + { +struct argo_ring_info *ring_info, *next; +struct list_head *bucket = &d->argo->ring_hash[i]; + +list_for_each_entry_safe (ring_info, next, bucket, node)
list_first_entry_or_null and you can get rid of next.
Quoted 14 lines
xen/common/argo.c +14
-0+ring_remove_info (d, ring_info); + } + d->argo->ring_count =0 ; +} + +/* +* Tear down all rings of other domains where src_d domain is the partner. +* (ie. it is the single domain that can send to those rings.) +* This will also cancel any pending notifications about those rings. +*/ +static void +partner_rings_remove (struct domain *src_d) +{ +unsigned int i;
xen/common/argo.c line 504
Quoted 10 lines
xen/common/argo.c +10
-0+ +ASSERT (LOCKING_Write_L1); + +for ( i =0 ; i <ARGO_HASHTABLE_SIZE ; ++i ) + { +struct argo_send_info *send_info, *next; +struct list_head *bucket = &src_d->argo->send_hash[i]; + +/* Remove all ents from the send list. Take each off their ring list. */ +list_for_each_entry_safe (send_info, next, bucket, node)
list_first_entry_or_null and you can get rid of next.
Quoted 33 lines
xen/common/argo.c +28
-0+ { +struct domain *dst_d =get_domain_by_id (send_info->id.domain_id); + +if ( dst_d && dst_d->argo ) + { +struct argo_ring_info *ring_info = +find_ring_info (dst_d, &send_info->id); + +if ( ring_info ) + { +ring_remove_info (dst_d, ring_info); + dst_d->argo->ring_count--; + } +else +ASSERT_UNREACHABLE (); + } +else +ASSERT_UNREACHABLE (); + +if ( dst_d ) +put_domain (dst_d); + +list_del (&send_info->node); +xfree (send_info); + } + } +} +long do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1,XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3,unsigned long arg4) {
xen/common/argo.c line 546
Quoted 10 lines
xen/common/argo.c +9
-1-return -ENOSYS ; +long rc = -EFAULT ; + +argo_dprintk ("->do_argo_op(%u,%p,%p,%lu,0x%lx)\n" , cmd, + (void *)arg1.p, (void *)arg2.p, arg3, arg4); + +if (unlikely (!opt_argo) ) +return -EOPNOTSUPP ; + + switch (cmd)
^ ^ missing spaces
Quoted 16 lines
xen/common/argo.c +9
-0+ { +default : + rc = -EOPNOTSUPP ; +break ; + } + +argo_dprintk ("<-do_argo_op(%u)=%ld\n" , cmd, rc); + +return rc; }#ifdef CONFIG_COMPAT @@ -42,6 +561,113 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, XEN_GUEST_HANDLE_PARAM(void) arg2, unsigned long arg3, unsigned long arg4) {
xen/common/argo.c compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 572
Quoted 10 lines
xen/common/argo.c +9
-1-return -ENOSYS ; +long rc = -EFAULT ; + +argo_dprintk ("->compat_argo_op(%u,%p,%p,%lu,0x%lx)\n" , cmd, + (void *)arg1.p, (void *)arg2.p, arg3, arg4); + +if (unlikely (!opt_argo) ) +return -EOPNOTSUPP ; + + switch (cmd)
^ ^ missing spaces
[...]
Quoted 22 lines
xen/common/domain.c +3
-0@@ -32,6 +32,7 @@#include <xen/grant_table.h> #include <xen/xenoprof.h> #include <xen/irq.h> +#include <xen/argo.h> #include <asm/debugger.h> #include <asm/p2m.h> #include <asm/processor.h> @@ -277,6 +278,8 @@ static void _domain_destroy(struct domain *d)xfree (d->pbuf); +argo_destroy (d); +rangeset_domain_destroy (d);free_cpumask_var (d->dirty_cpumask);
xen/common/domain.c struct domain *domain_create(domid_t domid, · line 451
Quoted 4 lines
xen/common/domain.c +1
-0@@ -445,6 +448,9 @@ struct domain *domain_create(domid_t domid, goto fail; init_status |= INIT_gnttab; +if ( (err =argo_init (d)) !=0 )
Here you are adding code that performs a variable assignment inside of a condition, much like what would be required to use list_first_entry_or_null.
Roger.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 04/15] argo: init, destroy and soft-reset, with enable command line opt
Quoted 18 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 15:49, <[email protected]> wrote:
xen/common/argo.c line 324
Quoted 12 lines · Christopher Clark
xen/common/argo.c +12
-0+static struct argo_ring_info * +find_ring_info (const struct domain *d,const struct argo_ring_id *id) +{ +struct argo_ring_info *ring_info; +const struct list_head *bucket; + +ASSERT (LOCKING_Read_rings_L2 (d)); + +/* List is not modified here. Search and return the match if found. */ + bucket = &d->argo->ring_hash[hash_index (id)]; + +list_for_each_entry (ring_info, bucket, node)I'm not sure what's the policy regarding
list_macros, should spaces be added between the parentheses?
Quoted 3 lines
list_for_each_entry(ring_info, bucket, node )I don't have a strong opinion either way.
We use a mix, because it's not really clear whether to consider such "kind-of-keywords".
Quoted 11 lines
xen/common/argo.c line 396
Quoted 9 lines
xen/common/argo.c +9
-0+static void +pending_remove_all (const struct domain *d,struct argo_ring_info *ring_info) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* Delete all pending notifications from this ring's list. */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node)
list_first_entry_or_nulland you can get rid of next.
Well, that's a substitution I wouldn't like to do while committing, as it means doing a little more than a simple text replacement.
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 04/15] argo: init, destroy and soft-reset, with enable command line opt
Quoted 37 lines · Jan Beulich
Quoted 18 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 15:49, <[email protected]> wrote:
xen/common/argo.c line 324
Quoted 12 lines · Christopher Clark
xen/common/argo.c +12
-0+static struct argo_ring_info * +find_ring_info (const struct domain *d,const struct argo_ring_id *id) +{ +struct argo_ring_info *ring_info; +const struct list_head *bucket; + +ASSERT (LOCKING_Read_rings_L2 (d)); + +/* List is not modified here. Search and return the match if found. */ + bucket = &d->argo->ring_hash[hash_index (id)]; + +list_for_each_entry (ring_info, bucket, node)I'm not sure what's the policy regarding
list_macros, should spaces be added between the parentheses?Quoted 3 lines
list_for_each_entry(ring_info, bucket, node )I don't have a strong opinion either way.
We use a mix, because it's not really clear whether to consider such "kind-of-keywords".
Quoted 11 lines
xen/common/argo.c line 396
Quoted 9 lines
xen/common/argo.c +9
-0+static void +pending_remove_all (const struct domain *d,struct argo_ring_info *ring_info) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* Delete all pending notifications from this ring's list. */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node)
list_first_entry_or_nulland you can get rid of next.
Quoted 2 lines
Well, that's a substitution I wouldn't like to do while committing, as it means doing a little more than a simple text replacement.
From a functionality PoV the code does exactly the same, I just think list_first_entry_or_null is more suitable and I haven't been convinced by the argument posted on v5.
However since this is argos specific code I don't have such a strong opinion, and it can always be improved later by a followup patch. My RB stands despite the usage of list_for_each_entry_safe.
Roger.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 04/15] argo: init, destroy and soft-reset, with enable command line opt
Quoted 16 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 17:05, <[email protected]> wrote:
Quoted 14 lines · Jan Beulich
Quoted 13 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 15:49, <[email protected]> wrote:
xen/common/argo.c line 396
Quoted 9 lines · Christopher Clark
xen/common/argo.c +9
-0+static void +pending_remove_all (const struct domain *d,struct argo_ring_info *ring_info) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* Delete all pending notifications from this ring's list. */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node)
list_first_entry_or_nulland you can get rid of next.
Quoted 6 lines
Quoted 2 lines
Well, that's a substitution I wouldn't like to do while committing, as it means doing a little more than a simple text replacement.
From a functionality PoV the code does exactly the same, I just think
list_first_entry_or_nullis more suitable and I haven't been convinced by the argument posted on v5.
Well, we're in agreement. All I did say is that a change like this I'd like to see done by the submitter, not the committer (at least not if I end up being the latter).
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 04/15] argo: init, destroy and soft-reset, with enable command line opt
Quoted 51 lines · Roger Pau Monné
Quoted 49 lines · Christopher Clark
Initialises basic data structures and performs teardown of argo state for domain shutdown.
Inclusion of the Argo implementation is dependent on
CONFIG_ARGO.Introduces a new Xen command line parameter 'argo': bool to enable/disable the argo hypercall. Defaults to disabled.
New headers:
public/argo.h: with definions of addresses and ring structure, including indexes for atomic update for communication between domain and hypervisor.
xen/argo.h: to expose the hooks for integration into domain lifecycle:
argo_init: per-domain init of argo data structures fordomain_create.
argo_destroy: teardown fordomain_destroyand the error exit
path ofdomain_create.
argo_soft_reset: reset of domain state fordomain_soft_reset.Adds a new field to struct domain: struct
argo_domain*argo;In accordance with recent work on
_domain_destroy,argo_destroyis idempotent. It will tear down: all rings registered by this domain, all rings where this domain is the single sender (ie. specified partner, non-wildcard rings), and all pending notifications where this domain is awaiting signal about available space in the rings of other domains.A count will be maintained of the number of rings that a domain has registered in order to limit it below the fixed maximum limit defined here.
Macros are defined to verify the internal locking state within the argo implementation. The macros are ASSERTed on entry to functions to validate and document the required lock state prior to calling.
The hash function for the hashtables that hold ring state is derived from the string hashing function djb2 (http://www.cse.yorku.ca/~oz/hash.html)
by Daniel J. Bernstein. Basic testing with a limited number of domains and ports has shown reasonable distribution for the table size.The software license on the public header is the BSD license, standard procedure for the public Xen headers. The public header was originally posted under a GPL license at: [1]:
https://lists.xenproject.org/archives/html/xen-devel/2013-05/msg02710.htmlThe following ACK by Lars Kurth is to confirm that only people being employees of Citrix contributed to the header files in the series posted at
[1] and that thus the copyright of the files in question is fully owned by Citrix. The ACK also confirms that Citrix is happy for the header files to be published under a BSD license in this series (which is based on [1]).
Quoted 6 lines
Quoted 4 lines
Signed-off-by Christopher Clark <[email protected]>Acked-by Lars Kurth <[email protected]>Reviewed-by Ross Philipson <[email protected]>Tested-by Chris Patterson <[email protected]>Reviewed-by Roger Pau Monné <[email protected]>
ack
Quoted 2 lines
I have some comments below that could be fixed upon commit if the committer agrees.
I've applied all the requested changes below for the next series submission.
Quoted 22 lines
xen/common/argo.c line 324
Quoted 13 lines
xen/common/argo.c +12
-0+static struct argo_ring_info * +find_ring_info (const struct domain *d,const struct argo_ring_id *id) +{ +struct argo_ring_info *ring_info; +const struct list_head *bucket; + +ASSERT (LOCKING_Read_rings_L2 (d)); + +/* List is not modified here. Search and return the match if found. */ + bucket = &d->argo->ring_hash[hash_index (id)]; + +list_for_each_entry (ring_info, bucket, node)I'm not sure what's the policy regarding
list_macros, should spaces be added between the parentheses?
list_for_each_entry(ring_info, bucket, node )I don't have a strong opinion either way.
[...]
Quoted 11 lines
xen/common/argo.c line 396
Quoted 9 lines
xen/common/argo.c +9
-0+static void +pending_remove_all (const struct domain *d,struct argo_ring_info *ring_info) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* Delete all pending notifications from this ring's list. */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node)
list_first_entry_or_nulland you can get rid of next.
ack
Quoted 13 lines
Quoted 13 lines
xen/common/argo.c +13
-0+ { +/* For wildcard rings, remove each from their wildcard list too. */ +if ( ring_info->id.partner_id ==XEN_ARGO_DOMID_ANY ) +wildcard_pending_list_remove (ent->domain_id, ent); +list_del (&ent->node); +xfree (ent); + } + ring_info->npending =0 ; +} + +static void +wildcard_rings_pending_remove (struct domain *d) +{
Quoted 8 lines
xen/common/argo.c line 415
Quoted 6 lines
xen/common/argo.c +6
-0+struct pending_ent *ent, *next; + +ASSERT (LOCKING_Write_L1); + +/* Delete all pending signals to the domain about wildcard rings. */ +list_for_each_entry_safe (ent, next, &d->argo->wildcard_pend_list, node)
list_first_entry_or_nulland you can get rid of next.
ack
Quoted 16 lines
Quoted 15 lines
xen/common/argo.c +14
-0+ { +/* +* The ent->node deleted here, and the npending value decreased, +* belong to the ring_info of another domain, which is why this +* function requires holding W(L1): +* it implies the L3 lock that protects that ring_info struct. +*/ + ent->ring_info->npending--; +list_del (&ent->node); +list_del (&ent->wildcard_node); +xfree (ent); + } +} +
Quoted 16 lines
Quoted 4 lines
xen/common/argo.c +4
-0+ +static void +domain_rings_remove_all (struct domain *d) +{xen/common/argo.c line 480
Quoted 10 lines
xen/common/argo.c +10
-0+unsigned int i; + +ASSERT (LOCKING_Write_rings_L2 (d)); + +for ( i =0 ; i <ARGO_HASHTABLE_SIZE ; ++i ) + { +struct argo_ring_info *ring_info, *next; +struct list_head *bucket = &d->argo->ring_hash[i]; + +list_for_each_entry_safe (ring_info, next, bucket, node)
list_first_entry_or_nulland you can get rid of next.
ack
Quoted 18 lines
Quoted 18 lines
xen/common/argo.c +18
-0+ring_remove_info (d, ring_info); + } + d->argo->ring_count =0 ; +} + +/* +* Tear down all rings of other domains where src_d domain is the partner. +* (ie. it is the single domain that can send to those rings.) +* This will also cancel any pending notifications about those rings. +*/ +static void +partner_rings_remove (struct domain *src_d) +{ +unsigned int i; + +ASSERT (LOCKING_Write_L1); + +for ( i =0 ; i <ARGO_HASHTABLE_SIZE ; ++i )
Quoted 8 lines
xen/common/argo.c line 504
Quoted 6 lines
xen/common/argo.c +6
-0+ { +struct argo_send_info *send_info, *next; +struct list_head *bucket = &src_d->argo->send_hash[i]; + +/* Remove all ents from the send list. Take each off their ring list. */ +list_for_each_entry_safe (send_info, next, bucket, node)
list_first_entry_or_nulland you can get rid of next.
ack
Quoted 36 lines
Quoted 36 lines
xen/common/argo.c +30
-1+ { +struct domain *dst_d =get_domain_by_id (send_info->id.domain_id); + +if ( dst_d && dst_d->argo ) + { +struct argo_ring_info *ring_info = +find_ring_info (dst_d, &send_info->id); + +if ( ring_info ) + { +ring_remove_info (dst_d, ring_info); + dst_d->argo->ring_count--; + } +else +ASSERT_UNREACHABLE (); + } +else +ASSERT_UNREACHABLE (); + +if ( dst_d ) +put_domain (dst_d); + +list_del (&send_info->node); +xfree (send_info); + } + } +} +long do_argo_op (unsigned int cmd,XEN_GUEST_HANDLE_PARAM (void ) arg1,XEN_GUEST_HANDLE_PARAM (void ) arg2,unsigned long arg3,unsigned long arg4) { -return -ENOSYS ; +long rc = -EFAULT ; +
Quoted 8 lines
xen/common/argo.c line 546
Quoted 7 lines
xen/common/argo.c +7
-0+argo_dprintk ("->do_argo_op(%u,%p,%p,%lu,0x%lx)\n" , cmd, + (void *)arg1.p, (void *)arg2.p, arg3, arg4); + +if (unlikely (!opt_argo) ) +return -EOPNOTSUPP ; + + switch (cmd)^ ^ missing spaces
ack
Quoted 11 lines
Quoted 11 lines
xen/common/argo.c +9
-0+ { +default : + rc = -EOPNOTSUPP ; +break ; + } + +argo_dprintk ("<-do_argo_op(%u)=%ld\n" , cmd, rc); + +return rc; }
Quoted 16 lines
Quoted 5 lines
#ifdef
CONFIG_COMPATxen/common/argo.c +0
-0@@ -42,6 +561,113 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, XEN_GUEST_HANDLE_PARAM(void) arg2, unsigned long arg3, unsigned long arg4) {xen/common/argo.c compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 572
Quoted 10 lines
xen/common/argo.c +9
-1-return -ENOSYS ; +long rc = -EFAULT ; + +argo_dprintk ("->compat_argo_op(%u,%p,%p,%lu,0x%lx)\n" , cmd, + (void *)arg1.p, (void *)arg2.p, arg3, arg4); + +if (unlikely (!opt_argo) ) +return -EOPNOTSUPP ; + + switch (cmd)^ ^ missing spaces
ack
Quoted 26 lines
[...]
Quoted 25 lines
xen/common/domain.c +3
-0@@ -32,6 +32,7 @@#include <xen/grant_table.h> #include <xen/xenoprof.h> #include <xen/irq.h> +#include <xen/argo.h> #include <asm/debugger.h> #include <asm/p2m.h> #include <asm/processor.h> @@ -277,6 +278,8 @@ static void _domain_destroy(struct domain *d)xfree (d->pbuf); +argo_destroy (d); +rangeset_domain_destroy (d);free_cpumask_var (d->dirty_cpumask);@@ -445,6 +448,9 @@ struct domain *domain_create(domid_t domid, goto fail; init_status |= INIT_gnttab;
Quoted 5 lines
xen/common/domain.c struct domain *domain_create(domid_t domid, · line 451
Quoted 1 line
xen/common/domain.c +1
-0+if ( (err =argo_init (d)) !=0 )Here you are adding code that performs a variable assignment inside of a condition, much like what would be required to use
list_first_entry_or_null.
Noted.
Quoted 20 lines · Jan Beulich
Quoted 20 lines
Quoted 2 lines
Quoted 2 lines · On 31.01.19 at 17:05, <[email protected]> wrote:
Quoted 16 lines · Jan Beulich
Quoted 13 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 15:49, <[email protected]> wrote:
xen/common/argo.c line 396
Quoted 9 lines · Christopher Clark
xen/common/argo.c +9
-0+static void +pending_remove_all (const struct domain *d,struct argo_ring_info *ring_info) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* Delete all pending notifications from this ring's list. */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node)
list_first_entry_or_nulland you can get rid of next.Well, that's a substitution I wouldn't like to do while committing, as it means doing a little more than a simple text replacement.
Quoted 7 lines
Quoted 3 lines
From a functionality PoV the code does exactly the same, I just think
list_first_entry_or_nullis more suitable and I haven't been convinced by the argument posted on v5.Well, we're in agreement. All I did say is that a change like this I'd like to see done by the submitter, not the committer (at least not if I end up being the latter).
Ack, changes are applied for next series submission.
Christopher
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 5/15 errno: add POSIX error codes EMSGSIZE, ECONNREFUSED to the ABI
In reply to Christopher ClarkEMSGSIZE: Argo's sendv operation will return EMSGSIZE when an excess amount of data, across all iovs, has been supplied, exceeding either the statically configured maximum size of a transmittable message, or the (variable) size of the ring registered by the destination domain.
ECONNREFUSED: Argo's register operation will return ECONNREFUSED if a ring is being registered to communicate with a specific remote domain that does exist but is not argo-enabled.
These codes are described by POSIX here:
http://pubs.opengroup.org/onlinepubs/9699919799/basedefs/errno.h.html
EMSGSIZE : "Message too large"
ECONNREFUSED : "Connection refused".
The numeric values assigned to each are taken from Linux, as is the case for the existing error codes.
EMSGSIZE : 90
ECONNREFUSED : 111
--- xen/include/public/errno.h | 2 ++ 1 file changed, 2 insertions(+)
xen/include/public/errno.h +2 -0
@@ -102,6 +102,7 @@ XEN_ERRNO(EILSEQ, 84) /* Illegal byte sequence */ XEN_ERRNO(ERESTART, 85) /* Interrupted system call should be restarted */#endif XEN_ERRNO(ENOTSOCK, 88) /* Socket operation on non-socket */ +XEN_ERRNO(EMSGSIZE, 90) /* Message too large. */ XEN_ERRNO(EOPNOTSUPP, 95) /* Operation not supported on transport endpoint */ XEN_ERRNO(EADDRINUSE, 98) /* Address already in use */XEN_ERRNO (EADDRNOTAVAIL ,99 )/* Cannot assign requested address */
@@ -109,6 +110,7 @@ XEN_ERRNO(ENOBUFS, 105) /* No buffer space available */ XEN_ERRNO(EISCONN, 106) /* Transport endpoint is already connected */ XEN_ERRNO(ENOTCONN, 107) /* Transport endpoint is not connected */ XEN_ERRNO(ETIMEDOUT, 110) /* Connection timed out */ +XEN_ERRNO(ECONNREFUSED, 111) /* Connection refused */#undef XEN_ERRNO#endif /* XEN_ERRNO */
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 6/15 xen/arm: introduce guest_handle_for_field()
In reply to Christopher ClarkARM port of c/s bb544585: "introduce guest_handle_for_field()"
This helper turns a field of a GUEST_HANDLE into a GUEST_HANDLE.
---
v3: Added Stefano's Reviewed-by
v2: Added Paul's Reviewed-by
xen/include/asm-arm/guest_access.h | 3 +++ 1 file changed, 3 insertions(+)
xen/include/asm-arm/guest_access.h +3 -0
@@ -63,6 +63,9 @@ int access_guest_memory_by_ipa(struct domain *d, paddr_t ipa, void *buf, _y; \ }) +#define guest_handle_for_field (hnd, type, fld) \ + ((XEN_GUEST_HANDLE(type)) { &(hnd).p->fld }) +#define guest_handle_from_ptr (ptr, type) \ ((XEN_GUEST_HANDLE_PARAM(type)) { (type *)ptr })#define const_guest_handle_from_ptr (ptr, type) \
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 7/15 argo: implement the register op
In reply to Christopher ClarkThe register op is used by a domain to register a region of memory for receiving messages from either a specified other domain, or, if specifying a wildcard, any domain.
This operation creates a mapping within Xen's private address space that will remain resident for the lifetime of the ring. In subsequent commits, the hypervisor will use this mapping to copy data from a sending domain into this registered ring, making it accessible to the domain that registered the ring to receive data.
Wildcard any-sender rings are default disabled and registration will be refused with EPERM unless they have been specifically enabled with the new mac-permissive flag that is added to the argo boot option here. The reason why the default for wildcard rings is 'deny' is that there is currently no means to protect the ring from DoS by a noisy domain spamming the ring, affecting other domains ability to send to it. This will be addressed with XSM policy controls in subsequent work.
Since denying access to any-sender rings is a significant functional constraint, the new option "mac-permissive" for the argo bootparam enables overriding this. eg: "argo=1,mac-permissive=1"
The p2m type of the memory supplied by the guest for the ring must be p2m_ram_rw and the memory will be pinned as PGT_writable_page while the ring is registered.
This hypercall op and its interface currently only supports 4K-sized pages.
---
v6 #09 Jan: add compat ABI
v6 #07 Jan: add argo_message_header to xlat.lst and invoke the CHECK v6 #07 Jan: xlat.lst: move argo struct entries to alphabetical position v5 #07 Roger: add BUILD_BUG_ON for MAX_RING_SIZE, PAGE_SIZE
v5 #07 Roger: gprintk(XENLOG_ERR,.. for denied existing ring
v5: add compat validation macros to primary source file: common/argo.c v5 : convert hypercall arg structs to struct form for compat checking
v5: dropped external file for compat macros: common/compat/argo.c
v4 v3#07 Jan: shrink critical sections in register_ring
v4 v3#07 Jan: revise register flag MASK in header, note 32-bitness of args v4 feedback: use standard data structures per common code, not loop macros v4 Andrew: use the single argo command line option list
v4 #07 Jan: rewrite find_ring_mfn to use check_get_page_from_gfn
v4 #07 Roger: add FIXME to ring_map_page for vmap contiguous ring mapping v3 #07 Jan: comment: minimum ring size is based on minimum-sized message v3 #04 Andrew: reference CONFIG_ARGO in the command line documentation v3 #07 Jan: register_ring: fold else, if into else-if to drop indent v3 #07 Jan: remove no longer used guest_handle_is_aligned macros
v3 #07 Jan: remove dead code from find_ring_mfns
v3 #07 Jan: fix format string indention in printks
v3 #07 Jan: remove redundant bounds check on npage in find_ring_mfns v3 #08 self/Roger: improve dprintk output in find_ring_info like find_send_info v3 #07 Jan: rename ring_find_info to find_ring_info
v3 #07 Jan: use array_index_nospec in ring_map_page
v3 #07 Jan: fix numeric entries in printk format strings
v3 #7 Jan: drop unneeded parentheses from ROUNDUP_MESSAGE defn
v3 #10 Roger: move find functions to top of file and drop prototypes v3 #03 meld compat check for hypercall arg register struct
v3 #04 Roger/Jan: make lock names clearer and assert their state
v3 #04 Jan: port -> aport with type; distinguish argo port from evtchn v3 feedback #07 Eric: fix header max ring size comment units
v3 feedback #04 Roger: mfn_mapping: void* instead of uint8_t*
v3 use %u for printing unsigned ints in find_ring_mfns
v3 feedback #04 Jan: uint32_t -> unsigned int for npage in register_ring v3 feedback #04 Roger: drop npages struct member, calculate from len v3 : register_ring: uint32_t -> unsigned int for private_tx_ptr
v3 feedback Roger/Jan: ASSERT currd is current->domain or use 'd' variable name v3 feedback #07 Roger: use opt_argo_mac_permissive : a boolean opt v3 feedback #04 Roger: reorder #includes to alphabetical order
v3 feedback #07 Roger: drop comment re: Intel EPT/AMD NPT for write-only mapping v3 feedback #07 Roger: drop ptr arithmetic in update_tx_ptr, use ring struct cast v3 feedback #07 Roger: drop newline in ring_map_page
v3 feedback #07 Roger: drop unneeded null check before xfree
v3 feedback #07 Roger: use return and drop out label in register_ring v3 Stefano: add 4K page constraint to header file comment & commit msg v3 Julien/Stefano: 4K granularity ok: use 64-bit gfns in register interface v2 self: disallow ring resize via reregister
v2 feedback Jan: drop cookie, implement teardown
v2 feedback Jan: drop message from argo_message_op
v2 self: move hash_index function below locking comment
v2 self: OVERHAUL
v2 self/Jan: remove use of magic verification field and tidy up
v2 self: merge max and min ring size check clauses
v2 feedback v1#13 Roger: use OS-supplied roundup; drop from public header v2 feedback #9, Jan: use the argo-mac bootparam at point of introduction v2 feedback #9, Jan: rename boot opt variable to comply with convention v2 feedback #9, Jan: rename the argo_mac bootparam to argo-mac
v2 feedback #9 Jan: document argo boot opt in xen-command-line.markdown v1,2 feedback Jan/Roger/Paul: drop errno returning guest access functions v1 feedback Roger, Jan: drop argo prefix on static functions
v1 feedback Roger: s/pfn/gfn/ and retire always-64-bit type
v2. feedback Jan: document the argo-mac boot opt
v2. feedback Jan: simplify re-register, drop mappings
v1 #13 feedback Jan: revise use of guest_handle_okay vs __copy ops v1 #13 feedback, Jan: register op : s/ECONNREFUSED/ESRCH/
v1 #5 (#13) feedback Paul: register op: use currd in do_message_op v1 #13 feedback, Paul: register op: use mfn_eq comparator
v1 #5 (#13) feedback Paul: register op: use currd in argo_register_ring v1 #13 feedback Paul: register op: whitespace, unsigned, bounds check v1 #13 feedback Paul: use of hex in limit constant definition
v1 #13 feedback Paul, register op: set nmfns on loop termination
v1 #13 feedback Paul: register op: do/while -> gotos, reindent
v1 argo_ring_map_page: drop uint32_t for unsigned int
v1. #13 feedback Julien: use page descriptors instead of gpfns.
- adds ABI support for pages with different granularity.
v1 feedback #13, Paul: adjust log level of message
v1 feedback #13, Paul: use gprintk for guest-triggered warning
v1 feedback #13, Paul: gprintk and XENLOG_DEBUG for ring registration v1 feedback #13, Paul: use gprintk for errs in argo_ring_map_page
v1 feedback #13, Paul: use ENOMEM if global mapping fails
v1 feedback Paul: overflow check before shift
v1: add define for copy_field_to_guest_errno
v1: fix gprintk use for ARM as its defn dislikes split format strings
v1: use copy_field_to_guest_errno
v1 feedback #13, Jan: argo_hash_fn: no inline, rename, change type v1 feedback #13, Paul, Jan: EFAULT -> ENOMEM in argo_ring_map_page v1 feedback #13, Jan: rename page var in argo_ring_map_page
v1 feedback #13, Jan: switch uint8_t* to void* and drop cast
v1 feedback #13, Jan: switch memory barrier to smp_wmb
v1 feedback #13, Jan: make 'ring' comment comply with single-line style v1 feedback #13, Jan: use xzalloc_array, drop loop NULL init
v1 feedback #13, Jan: init bool with false rather than 0
v1 feedback #13 Jan: use __copy; define and use __copy_field_to_guest_errno v1 feedback #13, Jan: use xzalloc, drop individual init zeroes
v1 feedback #13, Jan: prefix public namespace with xen
v1 feedback #13, Jan: blank line after op case in do_argo_message_op v1 self: reflow comment in argo_ring_map_page to within 80 char len v1 feedback #13, Roger: use true not 1 in assign to update_tx_ptr bool v1 feedback #21, Jan: fold in the array_index_nospec hash function guards v1 feedback #18, Jan: fold the max ring count limit into the series v1 self: use unsigned long type for XEN_ARGO_REGISTER_FLAG_MASK
v1: feedback #15 Jan: handle upper-halves of hypercall args
v1. feedback #13 Jan: add comment re: page alignment
v1. self: confirm ring magic presence in supplied page array
v1. feedback #13 Jan: add comment re: minimum ring size
v1. feedback #13 Roger: use ASSERT_UNREACHABLE
v1. feedback Roger: add comment to hash function
docs/misc/xen-command-line.pandoc | 9 +- xen/common/argo.c | 527 +++++++++++++++++++++++++++++++++++++- xen/include/public/argo.h | 70 +++++ xen/include/xlat.lst | 2 + 4 files changed, 606 insertions(+), 2 deletions(-)
docs/misc/xen-command-line.pandoc +8 -1
@@ -183,7 +183,7 @@ in combination with cpuidle. This option is only expected to be useful for developers wishing Xen to fall back to older timing methods on newer hardware. ### argo - = List of [ <bool> ] + = List of [ <bool>, mac-permissive=<bool> ] Controls for the Argo hypervisor-mediated interdomain communication service.
@@ -195,6 +195,13 @@ point of authority. Guests may register memory rings to recieve messages, query the status of other domains, and send messages by hypercall, all subject to appropriate auditing by Xen. Argo is disabled by default. +* The `mac-permissive` boolean controls whether wildcard receive rings may be + registered (`mac-permissive=1`) or may not be registered + (`mac-permissive=0`). + + This option is disabled by default, to protect domains from a DoS by a + buggy or malicious other domain spamming the ring. + ### asid (x86) > `= <boolean>`
xen/common/argo.c +526 -1
593 lines not on this page · raw message
xen/include/public/argo.h +70 -0
80 lines not on this page · raw message
xen/include/xlat.lst +2 -0
@@ -31,7 +31,9 @@ ! mc_physcpuinfo arch-x86/xen-mca.h ? page_offline_action arch-x86/xen-mca.h ? argo_addr argo.h +? argo_register_ring argo.h ? argo_ring argo.h +? argo_ring_message_header argo.h ? evtchn_alloc_unbound event_channel.h ? evtchn_bind_interdomain event_channel.h ? evtchn_bind_ipi event_channel.h
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 07/15] argo: implement the register op
Quoted 22 lines · Christopher Clark
The register op is used by a domain to register a region of memory for receiving messages from either a specified other domain, or, if specifying a wildcard, any domain.
This operation creates a mapping within Xen's private address space that will remain resident for the lifetime of the ring. In subsequent commits, the hypervisor will use this mapping to copy data from a sending domain into this registered ring, making it accessible to the domain that registered the ring to receive data.
Wildcard any-sender rings are default disabled and registration will be refused with EPERM unless they have been specifically enabled with the new mac-permissive flag that is added to the argo boot option here. The reason why the default for wildcard rings is 'deny' is that there is currently no means to protect the ring from DoS by a noisy domain spamming the ring, affecting other domains ability to send to it. This will be addressed with XSM policy controls in subsequent work.
Since denying access to any-sender rings is a significant functional constraint, the new option "mac-permissive" for the argo bootparam enables overriding this. eg: "argo=1,mac-permissive=1"
Quoted 8 lines
The p2m type of the memory supplied by the guest for the ring must be
p2m_ram_rwand the memory will be pinned asPGT_writable_pagewhile the ring is registered.This hypercall op and its interface currently only supports 4K-sized pages.
Signed-off-by Christopher Clark <[email protected]>Tested-by Chris Patterson <[email protected]>
Just one style issue below that should be fixed before commit, and two comments:
Quoted 15 lines
xen/common/argo.c +15
-0+static int +ring_map_page (const struct domain *d,struct argo_ring_info *ring_info, +unsigned int i,void **out_ptr) +{ +ASSERT (LOCKING_L3 (d, ring_info)); + +/* +* FIXME: Investigate using vmap to create a single contiguous virtual +* address space mapping of the ring instead of using the array of single +* page mappings. +* Affects logic in memcpy_to_guest_ring, the mfn_mapping array data +* structure, and places where ring mappings are added or removed. +*/ + +if ( i >= ring_info->nmfns )
xen/common/argo.c ring_unmap(const struct domain *d, struct argo_ring_info *ring_info) · line 415
Quoted 10 lines
xen/common/argo.c +10
-0+ { +gprintk (XENLOG_ERR , +"argo: ring (vm%u:%x vm%u) %p attempted to map page %u of %u\n" , + ring_info->id.domain_id, ring_info->id.aport, + ring_info->id.partner_id, ring_info, i, ring_info->nmfns); +return -ENOMEM ; + } + i =array_index_nospec (i, ring_info->nmfns); + +if ( !ring_info->mfns || !ring_info->mfn_mapping)
^ missing space
[...]
Quoted 7 lines
xen/common/argo.c +7
-0+static int +copy_gfn_from_handle (XEN_GUEST_HANDLE_PARAM (void ) gfn_hnd,bool compat, +unsigned int i,gfn_t *out_gfn) +{ +int ret; + +#ifdef CONFIG_COMPAT
xen/common/argo.c partner_rings_remove(struct domain *src_d) · line 666
Quoted 10 lines
xen/common/argo.c +10
-0+if ( compat ) + { +XEN_GUEST_HANDLE_PARAM (compat_pfn_t) c_gfn_hnd = +guest_handle_cast (gfn_hnd, compat_pfn_t); +compat_pfn_t c_gfn; + + ret =__copy_from_guest_offset (&c_gfn, c_gfn_hnd, i,1 ) ? -EFAULT :0 ; + *out_gfn =_gfn (c_gfn); + } +else
AFAICT you could place the #endif here and avoid the one below.
xen/common/argo.c compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 582
Quoted 5 lines
xen/common/argo.c +0
-1@@ -579,7 +1105,6 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,argo_dprintk ("<-compat_argo_op(%u)=%ld\n" , cmd, rc);return rc; -
This looks like a stray fix that should have gone into a different patch.
Roger.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 07/15] argo: implement the register op
Quoted 30 lines · Roger Pau Monné
Quoted 28 lines · Christopher Clark
The register op is used by a domain to register a region of memory for receiving messages from either a specified other domain, or, if specifying a wildcard, any domain.
This operation creates a mapping within Xen's private address space that will remain resident for the lifetime of the ring. In subsequent commits, the hypervisor will use this mapping to copy data from a sending domain into this registered ring, making it accessible to the domain that registered the ring to receive data.
Wildcard any-sender rings are default disabled and registration will be refused with EPERM unless they have been specifically enabled with the new mac-permissive flag that is added to the argo boot option here. The reason why the default for wildcard rings is 'deny' is that there is currently no means to protect the ring from DoS by a noisy domain spamming the ring, affecting other domains ability to send to it. This will be addressed with XSM policy controls in subsequent work.
Since denying access to any-sender rings is a significant functional constraint, the new option "mac-permissive" for the argo bootparam enables overriding this. eg: "argo=1,mac-permissive=1"
The p2m type of the memory supplied by the guest for the ring must be
p2m_ram_rwand the memory will be pinned asPGT_writable_pagewhile the ring is registered.This hypercall op and its interface currently only supports 4K-sized pages.
Quoted 7 lines
Quoted 2 lines
Signed-off-by Christopher Clark <[email protected]>Tested-by Chris Patterson <[email protected]>Just one style issue below that should be fixed before commit, and two comments:
Reviewed-by Roger Pau Monné <[email protected]>
Thanks
Quoted 18 lines
Quoted 18 lines
xen/common/argo.c +18
-0+static int +ring_map_page (const struct domain *d,struct argo_ring_info *ring_info, +unsigned int i,void **out_ptr) +{ +ASSERT (LOCKING_L3 (d, ring_info)); + +/* +* FIXME: Investigate using vmap to create a single contiguous virtual +* address space mapping of the ring instead of using the array of single +* page mappings. +* Affects logic in memcpy_to_guest_ring, the mfn_mapping array data +* structure, and places where ring mappings are added or removed. +*/ + +if ( i >= ring_info->nmfns ) + { + gprintk(XENLOG_ERR, + "argo: ring (vm%u:%x vm%u) %p attempted to map page %u of %u\n",
Quoted 8 lines
xen/common/argo.c ring_unmap(const struct domain *d, struct argo_ring_info *ring_info) · line 415
Quoted 7 lines
xen/common/argo.c +7
-0+ ring_info->id.domain_id, ring_info->id.aport, + ring_info->id.partner_id, ring_info, i, ring_info->nmfns); +return -ENOMEM ; + } + i =array_index_nospec (i, ring_info->nmfns); + +if ( !ring_info->mfns || !ring_info->mfn_mapping)^ missing space
ack
Quoted 12 lines
[...]
Quoted 11 lines
xen/common/argo.c +11
-0+static int +copy_gfn_from_handle (XEN_GUEST_HANDLE_PARAM (void ) gfn_hnd,bool compat, +unsigned int i,gfn_t *out_gfn) +{ +int ret; + +#ifdef CONFIG_COMPAT +if ( compat ) + { +XEN_GUEST_HANDLE_PARAM (compat_pfn_t) c_gfn_hnd = +guest_handle_cast (gfn_hnd, compat_pfn_t);
Quoted 8 lines
xen/common/argo.c partner_rings_remove(struct domain *src_d) · line 666
Quoted 6 lines
xen/common/argo.c +6
-0+compat_pfn_t c_gfn; + + ret =__copy_from_guest_offset (&c_gfn, c_gfn_hnd, i,1 ) ? -EFAULT :0 ; + *out_gfn =_gfn (c_gfn); + } +else AFAICT you could place the #endif here and avoid the one below.
ack, thanks.
Quoted 8 lines
xen/common/argo.c compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 582
Quoted 5 lines
xen/common/argo.c +0
-1@@ -579,7 +1105,6 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,argo_dprintk ("<-compat_argo_op(%u)=%ld\n" , cmd, rc);return rc; -This looks like a stray fix that should have gone into a different patch.
ack, fixed.
Christopher
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 8/15 argo: implement the unregister op
In reply to Christopher ClarkTakes a single argument: a handle to the ring unregistration struct, which specifies the port and partner domain id or wildcard.
The ring's entry is removed from the hashtable of registered rings; any entries for pending notifications are removed; and the ring is unmapped from Xen's address space.
If the ring had been registered to communicate with a single specified domain (ie. a non-wildcard ring) then the partner domain state is removed from the partner domain's argo send_info hash table.
---
v6 #09 Jan: add compat ABI
v6 Chris: add Tested-by
v6 Roger: add Reviewed-by
v6 #04 Jan: xlat.lst: move argo struct entries to alphabetical position v6 #04 Roger: use list_for_each_entry for looping
v5: add compat validation macros to primary source file: common/argo.c
v5: dropped external file for compat macros: common/compat/argo.c v4 # Jan: shrink the critical sections in unregister
v4 : use standard data structures as per common code
v4 #08 Roger: skip send_info lookup for wildcard rings
v4: add ASSERT_UNREACHABLE for missing sender domain or send_info
v4: reduce indentation by using goto
v4: add unlikely to currd->argo check
v4 #08 Jan: move put_domain outside L2 critical section
v4: include ring data in debug output when ring not found
v3 #08 Jan: pull xfree out of exclusive critical sections in unregister_ring v3 #08 Jan: rename send_find_info to find_send_info
v3 #07 Jan: rename ring_find_info to find_ring_info
v3 #08 Roger: use return and remove the out label in unregister_ring v3 #08 Roger: better debug output in send_find_info
v3 #10 Roger: move find functions to top of file and drop prototypes v3 #04 Jan: meld compat check for unregister_ring struct
v3 #04 Roger/Jan: make lock names clearer and assert their state v3 #04 Jan: port -> aport with type; distinguish argo port from evtchn v3 feedback Roger/Jan: ASSERT currd is current->domain or use 'd' variable name v3 feedback #07 Roger: const the argo_ring_id structs in send_find_info v2 feedback Jan: drop cookie, implement teardown
v2 feedback Jan: drop message from argo_message_op
v2 self: OVERHAUL
v2 self: reorder logic to shorten critical section
v1 #13 feedback Jan: revise use of guest_handle_okay vs __copy ops v1 feedback Roger, Jan: drop argo prefix on static functions
v1,2 feedback Jan/Roger/Paul: drop errno returning guest access functions v1 #5 (#14) feedback Paul: use currd in do_argo_message_op
v1 #5 (#14) feedback Paul: full use currd in argo_unregister_ring v1 #13 (#14) feedback Paul: replace do/while with goto; reindent v1 self: add blank lines in unregister case in do_argo_message_op
v1: #13 feedback Jan: public namespace: prefix with xen
v1: #13 feedback Jan: blank line after op case in do_argo_message_op
v1: #14 feedback Jan: replace domain id override with validation
v1: #18 feedback Jan: meld the ring count limit into the series
v1: feedback #15 Jan: verify zero in unused hypercall args
xen/common/argo.c | 140 ++++++++++++++++++++++++++++++++++++++++++++++ xen/include/public/argo.h | 19 +++++++ xen/include/xlat.lst | 1 + 3 files changed, 160 insertions(+)
xen/common/argo.c +140 -0
@@ -37,6 +37,7 @@ CHECK_argo_addr; CHECK_argo_register_ring; CHECK_argo_ring; CHECK_argo_ring_message_header; +CHECK_argo_unregister_ring;#endif #define MAX_RINGS_PER_DOMAIN 128U
@@ -52,6 +53,7 @@ CHECK_argo_ring_message_header;DEFINE_XEN_GUEST_HANDLE (xen_argo_addr_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_register_ring_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_ring_t); +DEFINE_XEN_GUEST_HANDLE (xen_argo_unregister_ring_t);#ifdef CONFIG_COMPAT DEFINE_XEN_GUEST_HANDLE (compat_pfn_t);#endif
@@ -362,6 +364,36 @@ find_ring_info(const struct domain *d, const struct argo_ring_id *id)return NULL ; } +static struct argo_send_info * +find_send_info (const struct domain *d,const struct argo_ring_id *id) +{ +struct argo_send_info *send_info; +const struct list_head *bucket; + +ASSERT (LOCKING_send_L2 (d)); + +/* List is not modified here. Search and return the match if found. */ + bucket = &d->argo->send_hash[hash_index (id)]; + +list_for_each_entry (send_info, bucket, node) + { +const struct argo_ring_id *cmpid = &send_info->id; + +if ( cmpid->aport == id->aport && + cmpid->domain_id == id->domain_id && + cmpid->partner_id == id->partner_id ) + { +argo_dprintk ("found send_info for ring(%u:%x %u)\n" , + id->domain_id, id->aport, id->partner_id); +return send_info; + } + } +argo_dprintk ("no send_info for ring(%u:%x %u)\n" , + id->domain_id, id->aport, id->partner_id); + +return NULL ; +} +static void ring_unmap (const struct domain *d,struct argo_ring_info *ring_info) {
@@ -765,6 +797,85 @@ find_ring_mfns(struct domain *d, struct argo_ring_info *ring_info, }static long +unregister_ring (struct domain *currd, +XEN_GUEST_HANDLE_PARAM (xen_argo_unregister_ring_t) unreg_hnd) +{ +xen_argo_unregister_ring_t unreg; +struct argo_ring_id ring_id; +struct argo_ring_info *ring_info =NULL ; +struct argo_send_info *send_info =NULL ; +struct domain *dst_d =NULL ; + +ASSERT (currd == current->domain); + +if (copy_from_guest (&unreg, unreg_hnd,1 ) ) +return -EFAULT ; + +if ( unreg.pad ) +return -EINVAL ; + + ring_id.partner_id = unreg.partner_id; + ring_id.aport = unreg.aport; + ring_id.domain_id = currd->domain_id; + +read_lock (&L1_global_argo_rwlock); + +if (unlikely (!currd->argo) ) + { +read_unlock (&L1_global_argo_rwlock); +return -ENODEV ; + } + +write_lock (&currd->argo->rings_L2_rwlock); + + ring_info =find_ring_info (currd, &ring_id); +if ( !ring_info ) + goto out; + +ring_remove_info (currd, ring_info); + currd->argo->ring_count--; + +if ( ring_id.partner_id ==XEN_ARGO_DOMID_ANY ) + goto out; + + dst_d =get_domain_by_id (ring_id.partner_id); +if ( !dst_d || !dst_d->argo ) + { +ASSERT_UNREACHABLE (); + goto out; + } + +spin_lock (&dst_d->argo->send_L2_lock); + + send_info =find_send_info (dst_d, &ring_id); +if ( send_info ) +list_del (&send_info->node); +else +ASSERT_UNREACHABLE (); + +spin_unlock (&dst_d->argo->send_L2_lock); + + out: +write_unlock (&currd->argo->rings_L2_rwlock); + +read_unlock (&L1_global_argo_rwlock); + +if ( dst_d ) +put_domain (dst_d); + +xfree (send_info); + +if ( !ring_info ) + { +argo_dprintk ("unregister_ring: no ring_info found for ring(%u:%x %u)\n" , + ring_id.domain_id, ring_id.aport, ring_id.partner_id); +return -ENOENT ; + } + +return 0 ; +} + +static long register_ring(struct domain *currd, XEN_GUEST_HANDLE_PARAM(xen_argo_register_ring_t) reg_hnd, XEN_GUEST_HANDLE_PARAM(void) gfn_hnd,
@@ -1049,6 +1160,21 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_unregister_ring: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_unregister_ring_t) unreg_hnd = +guest_handle_cast (arg1, xen_argo_unregister_ring_t); + +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =unregister_ring (currd, unreg_hnd); +break ; + } +default : rc = -EOPNOTSUPP ;break ;
@@ -1096,6 +1222,20 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_unregister_ring: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_unregister_ring_t) unreg_hnd = +guest_handle_cast (arg1, xen_argo_unregister_ring_t); + +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =unregister_ring (currd, unreg_hnd); +break ; + }default : rc = -EOPNOTSUPP ;
xen/include/public/argo.h +19 -0
@@ -76,6 +76,13 @@ typedef struct xen_argo_register_ringuint32_t len; } xen_argo_register_ring_t; +typedef struct xen_argo_unregister_ring +{ +xen_argo_port_t aport; +domid_t partner_id; +uint16_t pad; +}xen_argo_unregister_ring_t ; +/* Messages on the ring are padded to a multiple of this size. */ #define XEN_ARGO_MSG_SLOT_SIZE 0x10
@@ -131,4 +138,16 @@ struct xen_argo_ring_message_header#define XEN_ARGO_REGISTER_FLAG_MASK XEN_ARGO_REGISTER_FLAG_FAIL_EXIST#endif +/* +* XEN_ARGO_OP_unregister_ring +* +* Unregister a previously-registered ring, ending communication. +* +* arg1: XEN_GUEST_HANDLE(xen_argo_unregister_ring_t) +* arg2: NULL +* arg3: 0 (ZERO) +* arg4: 0 (ZERO) +*/ +#define XEN_ARGO_OP_unregister_ring 2 +#endif
xen/include/xlat.lst +1 -0
@@ -34,6 +34,7 @@ ? argo_register_ring argo.h ? argo_ring argo.h ? argo_ring_message_header argo.h +? argo_unregister_ring argo.h ? evtchn_alloc_unbound event_channel.h ? evtchn_bind_interdomain event_channel.h ? evtchn_bind_ipi event_channel.h
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 9/15 argo: implement the sendv op; evtchn: expose send_guest_global_virq
In reply to Christopher Clarksendv operation is invoked to perform a synchronous send of buffers contained in iovs to a remote domain's registered ring.
It takes:
* A destination address (domid, port) for the ring to send to. It performs a most-specific match lookup, to allow for wildcard.
* A source address, used to inform the destination of where to reply.
* The address of an array of iovs containing the data to send
* .. and the length of that array of iovs
* and a 32-bit message type, available to communicate message context data (eg. kernel-to-kernel, separate from the application data).
If insufficient space exists in the destination ring, it will return -EAGAIN and Xen will notify the caller when sufficient space becomes available.
Accesses to the ring indices are appropriately atomic. The rings are mapped into Xen's private address space to write as needed and the mappings are retained for later use.
Notifications are sent to guests via VIRQ and send_guest_global_virq is exposed in the change to enable argo to call it. VIRQ_ARGO is claimed from the VIRQ previously reserved for this purpose (#11).
The VIRQ notification method is used rather than sending events using evtchn functions directly because:
* no current event channel type is an exact fit for the intended behaviour. ECS_IPI is closest, but it disallows migration to other VCPUs which is not necessarily a requirement for Argo.
* at the point of argo_init, allocation of an event channel is complicated by none of the guest VCPUs being initialized yet and the event channel logic expects that a valid event channel has a present VCPU.
* at the point of signalling a notification, the VIRQ logic is already
defensive: if d->vcpu[0] is NULL, the notification is just silently dropped, whereas the evtchn_send logic is not so defensive: vcpu[0] must not be NULL, otherwise a null pointer dereference occurs.
Using a VIRQ removes the need for the guest to query to determine which event channel notifications will be delivered on. This is also likely to simplify establishing future L0/L1 nested hypervisor argo communication.
---
v6 #09 Jan: introduce compat ABI
v6 : hoist read of iovs from guest from ringbuf_insert to do_argo_op v6 : simplify init of the NULL_hnd in ringbuf_insert
v6 #09 Jan: xlat.lst remove argo_iov as generated macro not in use v6 #04 Jan: xlat.lst: move argo struct entries to alphabetical position v6 #09 Roger: fix reference to VIRQ_ARGO in commit message
v6 #09 Roger: use list_for_each_entry for looping
v5 #09 Roger: add comment explaining post-iovs tx_ptr round up + wrap v5 #09 Roger: remove redundant len bounds check vs MAX_ARGO_MESSAGE_SIZE v5 #09 Roger: ringbuf_insert: WARN not ERR on empty iovs
v5 #09 Roger: bugfix: set rc = -EFAULT if !guest_handle_okay
v5 use EBUSY when cannot add to the pending ent queue: many domains active
v5: add compat validation macros to primary source file: common/argo.c
v5: dropped external file for compat macros: common/compat/argo.c
v5 : convert hypercall arg structs to struct form for compat checking v5 : switch argo_iov to needs translation in xlat.lst
v4 Jan: remove use of fixed-width types from iov_count, ringbuf_insert v4 #07 Jan: shrink critical sections in sendv
v3 #07 Jan: header: note 32-bitness of hypercall message tuype arg v4 : use standard data structures as per common code
v4 self: bugfix memcpy_to_guest_ring: head_len must check (offset + len) v4 #09 Roger: drop MESSAGE from VIRQ_ARGO_MESSAGE
v3 #07 Jan: rename ring_find_info* to find_ring_info*
v3 #07 Jan: fix numeric entries in printk format strings
v3 #10 Roger: move find functions to top of file and drop prototypes v3 #04 Jan: meld compat struct checking for hypercall args
v3 #04 Roger/Jan: make lock names clearer and assert their state
v3 #04 Jan: port -> aport with type; distinguish argo port from evtchn v3 feedback #09 Eric: fix len & offset sanity check in memcpy_to_guest_ring v3 feedback #04 Roger: newline fix in wildcard_pending_list_insert v3 feedback #04 Roger: drop npages struct member, calculate from len v3 #09 Roger: simplify EFAULT return in memcpy_to_guest_ring
v3 #09 Roger: add newline before return in get_sanitized_ring
v3 #09 Roger: replace while with for loop in iov_count
v3 #09 Roger: drop 0 in struct init in ringbuf_insert
v3 #09 Roger: comment for XEN_ARGO_MAXIOV: warn of stack overflow risk v3 #09 Roger: simplify while loop: for instead in ringbuf_insert
v3 #09 Roger: drop out label for returns in ringbuf_insert
v3 #09 Roger: drop newline in pending_queue
v3 #09 Roger: replace second goto label with error path unlock in sendv v3 #09 Jason: check iov_len vs MAX_ARGO_MESSAGE_SIZE in iov_count
v3 #09 Jason: check padding is zeroed in sendv op
v3 #09 Jason: memcpy_to_guest_ring: simpler code with better loop
v2 self: use ring_info backpointer in pending_ent to maintain npending v2 feedback Jan: drop cookie, implement teardown
v2 self: pending_queue: reap stale ents when in need of space
v2 self: pending_requeue: reclaim ents for stale domains
v2.feedback Jan: only override sender domid if DOMID_ANY
v2 feedback Jan: drop message from argo_message_op
v2 self: check npending vs maximum limit
v2 self: get_sanitized_ring instead of get_rx_ptr
v2 feedback v1#13 Jan: remove double read from ringbuf insert, lower MAX_IOV v2 self: make iov_count const
v2 self: iov_count : return EMSGSIZE for message too big
v2 self: OVERHAUL
v2 self: s/argo_pending_ent/pending_ent/g
v2 feedback v1#13 Roger: use OS-supplied roundup; drop from public header v1,2 feedback Jan/Roger/Paul: drop errno returning guest access functions v1 feedback Roger, Jan: drop argo prefix on static functions
v1 feedback #13 Jan: drop guest_handle_okay when using copy_from_guest
- reorder do_argo_op logic
v2 self: add _hnd suffix to iovs variable name to indicate guest handle type v2 self: replace use of XEN_GUEST_HANDLE_NULL with two existing macros v1 #15 feedback, Jan: sendv op : s/ECONNREFUSED/ESRCH/
v1 #5 (#15) feedback Paul: sendv: use currd in do_argo_message_op
v1 #13 (#15) feedback Paul: sendv op: do/while reindent only
v1 #13 (#15) feedback Paul: sendv op: do/while: argo_ringbuf_insert to goto style v1 #13 (#15) feedback Paul: sendv op: do/while: reindent only again v1 #13 (#15) feedback Paul: sendv op: do/while : goto
v1 #15 feedback Paul: sendv op: make page var: unsigned
v1 #15 feedback Paul: sendv op: new local var for PAGE_SIZE - offset v1 #8 feedback Jan: XEN_GUEST_HANDLE : C89 compliance
v1 rebase after switching register op from pfns to page descriptors v1 self: move iov DEFINE_XEN_GUEST_HANDLE out of public header into argo.c v1 #13 (#15) feedback Paul: fix loglevel for guest-triggered messages v1 : add compat xlat.lst entries
v1 self: switched notification to send_guest_global_virq instead of event
v1: fix gprintk use for ARM as its defn dislikes split format strings
v1: init len variable to satisfy ARM compiler initialized checking v1 #13 feedback Jan: rename page var
v1:#14 feedback Jan: uint8_t* -> void*
v1: #13 feedback Jan: public namespace: prefix with xen
v1: #13 feedback Jan: blank line after case op in do_argo_message_op
v1: #15 feedback Jan: add comments explaining why the writes don't overrun
v1: self: add ASSERT to support comment that overrun cannot happen
v1: self: fail on short writes where guest manipulated the iov_lens
v1: self: rename ent id to domain_id
v1: self: add moan for iov rewrite
v1. feedback #15 Jan: require the pad bits are zero
v1. feedback #15 Jan: drop NULL check in argo_signal_domain as now using VIRQ v1. self: store domain_cookie in pending ent
v1. feedback #15 Jan: use unsigned where possible
v1. feedback Jan: use handle type for iov_base in public iov interface v1. self: log whenever visible error occurs
v1 feedback #15, Jan: drop unnecessary mb
v1 self: only update internal tx_ptr if able to return success
and update the visible tx_ptr
v1 self: log on failure to map ring to update visible tx_ptr
v1 feedback #15 Jan: add comment re: notification size policy
v1 self/Roger? remove errant space after sizeof
v1. feedback #15 Jan: require iov pad be zero
v1. self: rename iov_base to iov_hnd for handle in public iov interface
v1: feedback #15 Jan: handle upper-halves of hypercall args; changes some
types in function signatures to match.
v1: self: add dprintk to sendv
v1: self: add debug output to argo_iov_count
v1. feedback #14 Jan: blank line before return in argo_iov_count
v1 feedback #15 Jan: verify src id, not override
xen/common/argo.c | 675 +++++++++++++++++++++++++++++++++++++++++++++ xen/common/event_channel.c | 2 +- xen/include/public/argo.h | 54 ++++ xen/include/public/xen.h | 2 +- xen/include/xen/event.h | 7 + xen/include/xlat.lst | 2 + 6 files changed, 740 insertions(+), 2 deletions(-)
xen/common/argo.c +675 -0
@@ -38,24 +38,33 @@ CHECK_argo_register_ring; CHECK_argo_ring; CHECK_argo_ring_message_header; CHECK_argo_unregister_ring; +CHECK_argo_send_addr;#endif #define MAX_RINGS_PER_DOMAIN 128U +#define MAX_PENDING_PER_RING 32U/* All messages on the ring are padded to a multiple of the slot size. */ #define ROUNDUP_MESSAGE (a) ROUNDUP((a), XEN_ARGO_MSG_SLOT_SIZE) +/* The maximum size of a message that may be sent on the largest Argo ring. */ +#define MAX_ARGO_MESSAGE_SIZE ((XEN_ARGO_MAX_RING_SIZE) - \ + (sizeof(struct xen_argo_ring_message_header)) - ROUNDUP_MESSAGE(1)) +/* Number of PAGEs needed to hold a ring of a given size in bytes */ #define NPAGES_RING (ring_len) \ (ROUNDUP((ROUNDUP_MESSAGE(ring_len) + sizeof(xen_argo_ring_t)), PAGE_SIZE) \ >> PAGE_SHIFT)DEFINE_XEN_GUEST_HANDLE (xen_argo_addr_t); +DEFINE_XEN_GUEST_HANDLE (xen_argo_iov_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_register_ring_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_ring_t); +DEFINE_XEN_GUEST_HANDLE (xen_argo_send_addr_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_unregister_ring_t);#ifdef CONFIG_COMPAT DEFINE_XEN_GUEST_HANDLE (compat_pfn_t); +DEFINE_COMPAT_HANDLE (compat_argo_iov_t);#endif static bool __read_mostly opt_argo;
@@ -364,6 +373,28 @@ find_ring_info(const struct domain *d, const struct argo_ring_id *id)return NULL ; } +static struct argo_ring_info * +find_ring_info_by_match (const struct domain *d,xen_argo_port_t aport, +domid_t partner_id) +{ +struct argo_ring_id id; +struct argo_ring_info *ring_info; + +ASSERT (LOCKING_Read_rings_L2 (d)); + + id.aport = aport; + id.domain_id = d->domain_id; + id.partner_id = partner_id; + + ring_info =find_ring_info (d, &id); +if ( ring_info ) +return ring_info; + + id.partner_id =XEN_ARGO_DOMID_ANY ; + +return find_ring_info (d, &id); +} +static struct argo_send_info *find_send_info (const struct domain *d,const struct argo_ring_id *id) {
@@ -395,6 +426,14 @@ find_send_info(const struct domain *d, const struct argo_ring_id *id) }static void +signal_domain (struct domain *d) +{ +argo_dprintk ("signalling domid:%u\n" , d->domain_id); + +send_guest_global_virq (d,VIRQ_ARGO ); +} + +static void ring_unmap (const struct domain *d,struct argo_ring_info *ring_info) {unsigned int i;
@@ -488,6 +527,387 @@ update_tx_ptr(const struct domain *d, struct argo_ring_info *ring_info,smp_wmb (); } +static int +memcpy_to_guest_ring (const struct domain *d,struct argo_ring_info *ring_info, +unsigned int offset, +const void *src,XEN_GUEST_HANDLE (uint8) src_hnd, +unsigned int len) +{ +unsigned int mfns_index = offset >>PAGE_SHIFT ; +void *dst; +int ret; +unsigned int src_offset =0 ; + +ASSERT (LOCKING_L3 (d, ring_info)); + + offset &= ~PAGE_MASK ; + +if ( len + offset >XEN_ARGO_MAX_RING_SIZE ) +return -EFAULT ; + +while ( len ) + { +unsigned int head_len = (offset + len) >PAGE_SIZE ?PAGE_SIZE - offset + : len; + + ret =ring_map_page (d, ring_info, mfns_index, &dst); +if ( ret ) +return ret; + +if ( src ) + { +memcpy (dst + offset, src + src_offset, head_len); + src_offset += head_len; + } +else + { +if (copy_from_guest (dst + offset, src_hnd, head_len) ) +return -EFAULT ; + +guest_handle_add_offset (src_hnd, head_len); + } + + mfns_index++; + len -= head_len; + offset =0 ; + } + +return 0 ; +} + +/* +* Use this with caution: rx_ptr is under guest control and may be bogus. +* See get_sanitized_ring for a safer alternative. +*/ +static int +get_rx_ptr (const struct domain *d,struct argo_ring_info *ring_info, +uint32_t *rx_ptr) +{ +void *src; +xen_argo_ring_t *ringp; +int ret; + +ASSERT (LOCKING_L3 (d, ring_info)); + +if ( !ring_info->nmfns || ring_info->nmfns <NPAGES_RING (ring_info->len) ) +return -EINVAL ; + + ret =ring_map_page (d, ring_info,0 , &src); +if ( ret ) +return ret; + + ringp = (xen_argo_ring_t *)src; + + *rx_ptr =read_atomic (&ringp->rx_ptr); + +return 0 ; +} + +/* +* get_sanitized_ring creates a modified copy of the ring pointers where +* the rx_ptr is rounded up to ensure it is aligned, and then ring +* wrap is handled. Simplifies safe use of the rx_ptr for available +* space calculation. +*/ +static int +get_sanitized_ring (const struct domain *d,xen_argo_ring_t *ring, +struct argo_ring_info *ring_info) +{ +uint32_t rx_ptr; +int ret; + +ASSERT (LOCKING_L3 (d, ring_info)); + + ret =get_rx_ptr (d, ring_info, &rx_ptr); +if ( ret ) +return ret; + + ring->tx_ptr = ring_info->tx_ptr; + + rx_ptr =ROUNDUP_MESSAGE (rx_ptr); +if ( rx_ptr >= ring_info->len ) + rx_ptr =0 ; + + ring->rx_ptr = rx_ptr; + +return 0 ; +} + +/* +* iov_count returns its count on success via an out variable to avoid +* potential for a negative return value to be used incorrectly +* (eg. coerced into an unsigned variable resulting in a large incorrect value) +*/ +static int +iov_count (const xen_argo_iov_t *piov,unsigned int niov, +unsigned int *count) +{ +unsigned int sum_iov_lens =0 ; + +if ( niov >XEN_ARGO_MAXIOV ) +return -EINVAL ; + +for ( ; niov--; piov++ ) + { +/* valid iovs must have the padding field set to zero */ +if ( piov->pad ) + { +argo_dprintk ("invalid iov: padding is not zero\n" ); +return -EINVAL ; + } + +/* check each to protect sum against integer overflow */ +if ( piov->iov_len >MAX_ARGO_MESSAGE_SIZE ) + { +argo_dprintk ("invalid iov_len: too big (%u)>%llu\n" , + piov->iov_len,MAX_ARGO_MESSAGE_SIZE ); +return -EINVAL ; + } + + sum_iov_lens += piov->iov_len; + +/* +* Again protect sum from integer overflow +* and ensure total msg size will be within bounds. +*/ +if ( sum_iov_lens >MAX_ARGO_MESSAGE_SIZE ) + { +argo_dprintk ("invalid iov series: total message too big\n" ); +return -EMSGSIZE ; + } + } + + *count = sum_iov_lens; + +return 0 ; +} + +static int +ringbuf_insert (const struct domain *d,struct argo_ring_info *ring_info, +const struct argo_ring_id *src_id,xen_argo_iov_t *iovs, +unsigned int niov,uint32_t message_type, +unsigned long *out_len) +{ +xen_argo_ring_t ring; +struct xen_argo_ring_message_header mh = { }; +int sp, ret; +unsigned int len =0 ; +xen_argo_iov_t *piov; +XEN_GUEST_HANDLE (uint8) NULL_hnd = { }; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* +* Obtain the total size of data to transmit -- sets the 'len' variable +* -- and sanity check that the iovs conform to size and number limits. +* Enforced below: no more than 'len' bytes of guest data +* (plus the message header) will be sent in this operation. +*/ + ret =iov_count (iovs, niov, &len); +if ( ret ) +return ret; + +/* +* Upper bound check the message len against the ring size. +* The message must not fill the ring; there must be at least one slot +* remaining so we can distinguish a full ring from an empty one. +* iov_count has already verified: len <= MAX_ARGO_MESSAGE_SIZE. +*/ +if ( (ROUNDUP_MESSAGE (len) +sizeof (struct xen_argo_ring_message_header )) + >= ring_info->len ) +return -EMSGSIZE ; + + ret =get_sanitized_ring (d, &ring, ring_info); +if ( ret ) +return ret; + +argo_dprintk ("ring.tx_ptr=%u ring.rx_ptr=%u ring len=%u" +" ring_info->tx_ptr=%u\n" , + ring.tx_ptr, ring.rx_ptr, ring_info->len, ring_info->tx_ptr); + +if ( ring.rx_ptr == ring.tx_ptr ) + sp = ring_info->len; +else + { + sp = ring.rx_ptr - ring.tx_ptr; +if ( sp <0 ) + sp += ring_info->len; + } + +/* +* Size bounds check against currently available space in the ring. +* Again: the message must not fill the ring leaving no space remaining. +*/ +if ( (ROUNDUP_MESSAGE (len) + +sizeof (struct xen_argo_ring_message_header )) >= sp ) + { +argo_dprintk ("EAGAIN\n" ); +return -EAGAIN ; + } + + mh.len = len +sizeof (struct xen_argo_ring_message_header ); + mh.source.aport = src_id->aport; + mh.source.domain_id = src_id->domain_id; + mh.message_type = message_type; + +/* +* For this copy to the guest ring, tx_ptr is always 16-byte aligned +* and the message header is 16 bytes long. +*/ +BUILD_BUG_ON ( +sizeof (struct xen_argo_ring_message_header ) !=ROUNDUP_MESSAGE (1 )); + +/* +* First data write into the destination ring: fixed size, message header. +* This cannot overrun because the available free space (value in 'sp') +* is checked above and must be at least this size. +*/ + ret =memcpy_to_guest_ring (d, ring_info, + ring.tx_ptr +sizeof (xen_argo_ring_t), + &mh, NULL_hnd,sizeof (mh)); +if ( ret ) + { +gprintk (XENLOG_ERR , +"argo: failed to write message header to ring (vm%u:%x vm%u)\n" , + ring_info->id.domain_id, ring_info->id.aport, + ring_info->id.partner_id); + +return ret; + } + + ring.tx_ptr +=sizeof (mh); +if ( ring.tx_ptr == ring_info->len ) + ring.tx_ptr =0 ; + +for ( piov = iovs; niov--; piov++ ) + { +XEN_GUEST_HANDLE (uint8) buf_hnd = piov->iov_hnd; +unsigned int iov_len = piov->iov_len; + +/* If no data is provided in this iov, moan and skip on to the next */ +if ( !iov_len ) + { +gprintk (XENLOG_WARNING , +"argo: no data iov_len=0 iov_hnd=%p ring (vm%u:%x vm%u)\n" , + buf_hnd.p, ring_info->id.domain_id, ring_info->id.aport, + ring_info->id.partner_id); + +continue ; + } + +if (unlikely (!guest_handle_okay (buf_hnd, iov_len)) ) + { +gprintk (XENLOG_ERR , +"argo: bad iov handle [%p, %u] (vm%u:%x vm%u)\n" , + buf_hnd.p, iov_len, + ring_info->id.domain_id, ring_info->id.aport, + ring_info->id.partner_id); + +return -EFAULT ; + } + + sp = ring_info->len - ring.tx_ptr; + +/* Check: iov data size versus free space at the tail of the ring */ +if ( iov_len > sp ) + { +/* +* Second possible data write: ring-tail-wrap-write. +* Populate the ring tail and update the internal tx_ptr to handle +* wrapping at the end of ring. +* Size of data written here: sp +* which is the exact full amount of free space available at the +* tail of the ring, so this cannot overrun. +*/ + ret =memcpy_to_guest_ring (d, ring_info, + ring.tx_ptr +sizeof (xen_argo_ring_t), +NULL , buf_hnd, sp); +if ( ret ) + { +gprintk (XENLOG_ERR , +"argo: failed to copy {%p, %d} (vm%u:%x vm%u)\n" , + buf_hnd.p, sp, + ring_info->id.domain_id, ring_info->id.aport, + ring_info->id.partner_id); + +return ret; + } + + ring.tx_ptr =0 ; + iov_len -= sp; +guest_handle_add_offset (buf_hnd, sp); + +ASSERT (iov_len <= ring_info->len); + } + +/* +* Third possible data write: all data remaining for this iov. +* Size of data written here: iov_len +* +* Case 1: if the ring-tail-wrap-write above was performed, then +* iov_len has been decreased by 'sp' and ring.tx_ptr is zero. +* +* We know from checking the result of iov_count: +* len + sizeof(message_header) <= ring_info->len +* We also know that len is the total of summing all iov_lens, so: +* iov_len <= len +* so by transitivity: +* iov_len <= len <= (ring_info->len - sizeof(msgheader)) +* and therefore: +* (iov_len + sizeof(msgheader) <= ring_info->len) && +* (ring.tx_ptr == 0) +* so this write cannot overrun here. +* +* Case 2: ring-tail-wrap-write above was not performed +* -> so iov_len is the guest-supplied value and: (iov_len <= sp) +* ie. less than available space at the tail of the ring: +* so this write cannot overrun. +*/ + ret =memcpy_to_guest_ring (d, ring_info, + ring.tx_ptr +sizeof (xen_argo_ring_t), +NULL , buf_hnd, iov_len); +if ( ret ) + { +gprintk (XENLOG_ERR , +"argo: failed to copy [%p, %u] (vm%u:%x vm%u)\n" , + buf_hnd.p, iov_len, ring_info->id.domain_id, + ring_info->id.aport, ring_info->id.partner_id); + +return ret; + } + + ring.tx_ptr += iov_len; + +if ( ring.tx_ptr == ring_info->len ) + ring.tx_ptr =0 ; + } + +/* +* Finished writing data from all iovs into the ring: now need to round up +* tx_ptr to align to the next message boundary, and then wrap if necessary. +*/ + ring.tx_ptr =ROUNDUP_MESSAGE (ring.tx_ptr); + +if ( ring.tx_ptr >= ring_info->len ) + ring.tx_ptr -= ring_info->len; + +update_tx_ptr (d, ring_info, ring.tx_ptr); + +/* +* At this point (and also on an error exit paths from this function) it is +* possible to unmap the ring_info, ie: +* ring_unmap(d, ring_info); +* but performance should be improved by not doing so, and retaining +* the mapping. +* An XSM policy control over level of confidentiality required +* versus performance cost could be added to decide that here. +*/ + + *out_len = len; + +return ret; +} +static void wildcard_pending_list_remove (domid_t domain_id,struct pending_ent *ent) {
@@ -508,6 +928,25 @@ wildcard_pending_list_remove(domid_t domain_id, struct pending_ent *ent) }static void +wildcard_pending_list_insert (domid_t domain_id,struct pending_ent *ent) +{ +struct domain *d =get_domain_by_id (domain_id); + +if ( !d ) +return ; + +ASSERT (LOCKING_Read_L1); + +if ( d->argo ) + { +spin_lock (&d->argo->wildcard_L2_lock); +list_add (&ent->wildcard_node, &d->argo->wildcard_pend_list); +spin_unlock (&d->argo->wildcard_L2_lock); + } +put_domain (d); +} + +static void pending_remove_all (const struct domain *d,struct argo_ring_info *ring_info) {struct pending_ent *ent, *next;
@@ -526,6 +965,66 @@ pending_remove_all(const struct domain *d, struct argo_ring_info *ring_info) ring_info->npending =0 ; } +static int +pending_queue (const struct domain *d,struct argo_ring_info *ring_info, +domid_t src_id,unsigned int len) +{ +struct pending_ent *ent; + +ASSERT (LOCKING_L3 (d, ring_info)); + +if ( ring_info->npending >=MAX_PENDING_PER_RING ) +return -EBUSY ; + + ent = xmalloc(struct pending_ent); +if ( !ent ) +return -ENOMEM ; + + ent->len = len; + ent->domain_id = src_id; + ent->ring_info = ring_info; + +if ( ring_info->id.partner_id ==XEN_ARGO_DOMID_ANY ) +wildcard_pending_list_insert (src_id, ent); +list_add (&ent->node, &ring_info->pending); + ring_info->npending++; + +return 0 ; +} + +static int +pending_requeue (const struct domain *d,struct argo_ring_info *ring_info, +domid_t src_id,unsigned int len) +{ +struct pending_ent *ent; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* List structure is not modified here. Update len in a match if found. */ +list_for_each_entry (ent, &ring_info->pending, node) + { +if ( ent->domain_id == src_id ) + { +/* +* Reuse an existing queue entry for a notification rather than add +* another. If the existing entry is waiting for a smaller size than +* the current message then adjust the record to wait for the +* current (larger) size to be available before triggering a +* notification. +* This assists the waiting sender by ensuring that whenever a +* notification is triggered, there is sufficient space available +* for (at least) any one of the messages awaiting transmission. +*/ +if ( ent->len < len ) + ent->len = len; + +return 0 ; + } + } + +return pending_queue (d, ring_info, src_id, len); +} +static void wildcard_rings_pending_remove (struct domain *d) {
@@ -1119,6 +1618,98 @@ register_ring(struct domain *currd,return ret; } +static long +sendv (struct domain *src_d,xen_argo_addr_t *src_addr, +const xen_argo_addr_t *dst_addr,xen_argo_iov_t *iovs,unsigned int niov, +uint32_t message_type) +{ +struct domain *dst_d =NULL ; +struct argo_ring_id src_id; +struct argo_ring_info *ring_info; +int ret =0 ; +unsigned long len =0 ; + +argo_dprintk ("sendv: (%u:%x)->(%u:%x) niov:%u type:%x\n" , + src_addr->domain_id, src_addr->aport, dst_addr->domain_id, + dst_addr->aport, niov, message_type); + +/* Check padding is zeroed. */ +if (unlikely (src_addr->pad || dst_addr->pad) ) +return -EINVAL ; + +if ( src_addr->domain_id ==XEN_ARGO_DOMID_ANY ) + src_addr->domain_id = src_d->domain_id; + +/* No domain is currently authorized to send on behalf of another */ +if (unlikely (src_addr->domain_id != src_d->domain_id) ) +return -EPERM ; + + src_id.aport = src_addr->aport; + src_id.domain_id = src_d->domain_id; + src_id.partner_id = dst_addr->domain_id; + + dst_d =get_domain_by_id (dst_addr->domain_id); +if ( !dst_d ) +return -ESRCH ; + +read_lock (&L1_global_argo_rwlock); + +if ( !src_d->argo ) + { + ret = -ENODEV ; + goto out_unlock; + } + +if ( !dst_d->argo ) + { +argo_dprintk ("!dst_d->argo, ECONNREFUSED\n" ); + ret = -ECONNREFUSED ; + goto out_unlock; + } + +read_lock (&dst_d->argo->rings_L2_rwlock); + + ring_info =find_ring_info_by_match (dst_d, dst_addr->aport, + src_id.domain_id); +if ( !ring_info ) + { +gprintk (XENLOG_ERR , +"argo: vm%u connection refused, src (vm%u:%x) dst (vm%u:%x)\n" , + current->domain->domain_id, src_id.domain_id, src_id.aport, + dst_addr->domain_id, dst_addr->aport); + + ret = -ECONNREFUSED ; + } +else + { +spin_lock (&ring_info->L3_lock); + + ret =ringbuf_insert (dst_d, ring_info, &src_id, iovs, niov, + message_type, &len); +if ( ret == -EAGAIN ) + { +argo_dprintk ("argo_ringbuf_sendv failed, EAGAIN\n" ); +/* requeue to issue a notification when space is there */ + ret =pending_requeue (dst_d, ring_info, src_id.domain_id, len); + } + +spin_unlock (&ring_info->L3_lock); + } + +read_unlock (&dst_d->argo->rings_L2_rwlock); + + out_unlock: +read_unlock (&L1_global_argo_rwlock); + +if ( ret >=0 ) +signal_domain (dst_d); + +if ( dst_d ) +put_domain (dst_d); + +return ( ret <0 ) ? ret : len; +} + long do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, XEN_GUEST_HANDLE_PARAM(void) arg2, unsigned long arg3,
@@ -1175,6 +1766,42 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_sendv: + { +xen_argo_send_addr_t send_addr; +xen_argo_iov_t iovs[XEN_ARGO_MAXIOV ]; +unsigned int niov; + +XEN_GUEST_HANDLE_PARAM (xen_argo_send_addr_t) send_addr_hnd = +guest_handle_cast (arg1, xen_argo_send_addr_t); +XEN_GUEST_HANDLE_PARAM (xen_argo_iov_t) iovs_hnd = +guest_handle_cast (arg2, xen_argo_iov_t); +/* arg3 is niov */ +/* arg4 is message_type. Must be a 32-bit value. */ + + rc =copy_from_guest (&send_addr, send_addr_hnd,1 ) ? -EFAULT :0 ; +if ( rc ) +break ; + +/* +* Reject niov above maximum limit or message_types that are outside +* 32 bit range. +*/ +if (unlikely ((arg3 >XEN_ARGO_MAXIOV ) || (arg4 & ~0xffffffffUL )) ) + { + rc = -EINVAL ; +break ; + } + niov =array_index_nospec (arg3,XEN_ARGO_MAXIOV +1 ); + + rc =copy_from_guest (iovs, iovs_hnd, niov) ? -EFAULT :0 ; +if ( rc ) +break ; + + rc =sendv (currd, &send_addr.src, &send_addr.dst, iovs, niov, arg4); +break ; + } +default : rc = -EOPNOTSUPP ;break ;
@@ -1237,6 +1864,54 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_sendv: + { +xen_argo_send_addr_t send_addr; +xen_argo_iov_t iovs[XEN_ARGO_MAXIOV ]; +unsigned int i; +unsigned int niov; + +XEN_GUEST_HANDLE_PARAM (xen_argo_send_addr_t) send_addr_hnd = +guest_handle_cast (arg1, xen_argo_send_addr_t); +/* arg2: iovs, arg3: niov, arg4: message_type */ + + rc =copy_from_guest (&send_addr, send_addr_hnd,1 ) ? -EFAULT :0 ; +if ( rc ) +break ; + +if (unlikely (arg3 >XEN_ARGO_MAXIOV ) ) + { + rc = -EINVAL ; +break ; + } + niov =array_index_nospec (arg3,XEN_ARGO_MAXIOV +1 ); + +/* +* Limited scope for compat_iovs array: enables a single copy_from_guest +* call and discards the array from the stack before calling sendv. +*/ + { +compat_argo_iov_t compat_iovs[XEN_ARGO_MAXIOV ]; + + rc =copy_from_guest (compat_iovs, arg2, niov) ? -EFAULT :0 ; +if ( rc ) +break ; + +for ( i =0 ; i < niov; i++ ) + { +#define XLAT_argo_iov_HNDL_iov_hnd (_d_, _s_) \ + guest_from_compat_handle((_d_)->iov_hnd, (_s_)->iov_hnd) + +XLAT_argo_iov (&iovs[i], &compat_iovs[i]); + +#undef XLAT_argo_iov_HNDL_iov_hnd + } + } + + rc =sendv (currd, &send_addr.src, &send_addr.dst, iovs, niov, arg4); +break ; + } +default : rc = -EOPNOTSUPP ;break ;
xen/common/event_channel.c +1 -1
@@ -746,7 +746,7 @@ void send_guest_vcpu_virq(struct vcpu *v, uint32_t virq)spin_unlock_irqrestore (&v->virq_lock, flags); } -static void send_guest_global_virq (struct domain *d,uint32_t virq) +void send_guest_global_virq (struct domain *d,uint32_t virq) {unsigned long flags;int port;
xen/include/public/argo.h +54 -0
70 lines not on this page · raw message
xen/include/public/xen.h +1 -1
@@ -178,7 +178,7 @@ DEFINE_XEN_GUEST_HANDLE(xen_ulong_t);#define VIRQ_CON_RING 8/* G. (DOM0) Bytes received on console */ #define VIRQ_PCPU_STATE 9/* G. (DOM0) PCPU state changed */ #define VIRQ_MEM_EVENT 10/* G. (DOM0) A memory event has occurred */ -#define VIRQ_XC_RESERVED 11/* G. Reserved for XenClient */ +#define VIRQ_ARGO 11/* G. Argo interdomain message notification */ #define VIRQ_ENOMEM 12/* G. (DOM0) Low on heap memory */ #define VIRQ_XENPMU 13/* V. PMC interrupt */
xen/include/xen/event.h +7 -0
@@ -29,6 +29,13 @@ void send_guest_vcpu_virq(struct vcpu *v, uint32_t virq);void send_global_virq (uint32_t virq);/* +* send_guest_global_virq: +* @d: Domain to which VIRQ should be sent +* @virq: Virtual IRQ number (VIRQ_*), must be global +*/ +void send_guest_global_virq (struct domain *d,uint32_t virq); + +/* * sent_global_virq_handler: Set a global VIRQ handler. * @d: New target domain for this VIRQ * @virq: Virtual IRQ number (VIRQ_*), must be global
xen/include/xlat.lst +2 -0
@@ -31,9 +31,11 @@ ! mc_physcpuinfo arch-x86/xen-mca.h ? page_offline_action arch-x86/xen-mca.h ? argo_addr argo.h +! argo_iov argo.h ? argo_register_ring argo.h ? argo_ring argo.h ? argo_ring_message_header argo.h +? argo_send_addr argo.h ? argo_unregister_ring argo.h ? evtchn_alloc_unbound event_channel.h ? evtchn_bind_interdomain event_channel.h
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 09/15] argo: implement the sendv op; evtchn: expose send_guest_global_virq
Quoted 41 lines · Christopher Clark
sendv operation is invoked to perform a synchronous send of buffers contained in iovs to a remote domain's registered ring.
It takes:
* A destination address (domid, port) for the ring to send to. It performs a most-specific match lookup, to allow for wildcard.
* A source address, used to inform the destination of where to reply.
* The address of an array of iovs containing the data to send
* .. and the length of that array of iovs
* and a 32-bit message type, available to communicate message context data (eg. kernel-to-kernel, separate from the application data).If insufficient space exists in the destination ring, it will return
-EAGAINand Xen will notify the caller when sufficient space becomes available.Accesses to the ring indices are appropriately atomic. The rings are mapped into Xen's private address space to write as needed and the mappings are retained for later use.
Notifications are sent to guests via VIRQ and
send_guest_global_virqis exposed in the change to enable argo to call it.VIRQ_ARGOis claimed from the VIRQ previously reserved for this purpose (#11).The VIRQ notification method is used rather than sending events using evtchn functions directly because:
* no current event channel type is an exact fit for the intended behaviour.
ECS_IPIis closest, but it disallows migration to other VCPUs which is not necessarily a requirement for Argo.* at the point of
argo_init, allocation of an event channel is complicated by none of the guest VCPUs being initialized yet and the event channel logic expects that a valid event channel has a present VCPU.* at the point of signalling a notification, the VIRQ logic is already
defensive: if d->vcpu[0] is NULL, the notification is just silently dropped, whereas theevtchn_sendlogic is not so defensive: vcpu[0] must not be NULL, otherwise a null pointer dereference occurs.
Quoted 6 lines
Using a VIRQ removes the need for the guest to query to determine which event channel notifications will be delivered on. This is also likely to simplify establishing future L0/L1 nested hypervisor argo communication.
Signed-off-by Christopher Clark <[email protected]>Tested-by Chris Patterson <[email protected]>
There's one style nit that I think can be fixed while committing:
Despite the usage of the open-coded mask below. As with previous patches this is argos code, so I'm not going to oppose, but again I think using such open coded masks is bad, and can lead to bugs in the code. It can be fixed by a follow up patch.
Quoted 22 lines
xen/common/argo.c +22
-0+static int +ringbuf_insert (const struct domain *d,struct argo_ring_info *ring_info, +const struct argo_ring_id *src_id,xen_argo_iov_t *iovs, +unsigned int niov,uint32_t message_type, +unsigned long *out_len) +{ +xen_argo_ring_t ring; +struct xen_argo_ring_message_header mh = { }; +int sp, ret; +unsigned int len =0 ; +xen_argo_iov_t *piov; +XEN_GUEST_HANDLE (uint8) NULL_hnd = { }; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* +* Obtain the total size of data to transmit -- sets the 'len' variable +* -- and sanity check that the iovs conform to size and number limits. +* Enforced below: no more than 'len' bytes of guest data +* (plus the message header) will be sent in this operation. +*/ + ret =iov_count (iovs, niov, &len);
xen/common/argo.c update_tx_ptr(const struct domain *d, struct argo_ring_info *ring_info, · line 716
Quoted 10 lines
xen/common/argo.c +10
-0+if ( ret ) +return ret; + +/* +* Upper bound check the message len against the ring size. +* The message must not fill the ring; there must be at least one slot +* remaining so we can distinguish a full ring from an empty one. +* iov_count has already verified: len <= MAX_ARGO_MESSAGE_SIZE. +*/ + if ( (ROUNDUP_MESSAGE(len) + sizeof(struct xen_argo_ring_message_header))
missing space ^
Quoted 1 line
+ >=
ring_info->len)
Align of >= also looks weird, should be aligned to the parenthesis before ROUNDUP_.
Quoted 16 lines
xen/common/argo.c +12
-0@@ -1175,6 +1766,42 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_sendv: + { +xen_argo_send_addr_t send_addr; +xen_argo_iov_t iovs[XEN_ARGO_MAXIOV ]; +unsigned int niov; + +XEN_GUEST_HANDLE_PARAM (xen_argo_send_addr_t) send_addr_hnd = +guest_handle_cast (arg1, xen_argo_send_addr_t); +XEN_GUEST_HANDLE_PARAM (xen_argo_iov_t) iovs_hnd = +guest_handle_cast (arg2, xen_argo_iov_t); +/* arg3 is niov */ +/* arg4 is message_type. Must be a 32-bit value. */
xen/common/argo.c do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 1790
Quoted 10 lines
xen/common/argo.c +10
-0+ + rc =copy_from_guest (&send_addr, send_addr_hnd,1 ) ? -EFAULT :0 ; +if ( rc ) +break ; + +/* +* Reject niov above maximum limit or message_types that are outside +* 32 bit range. +*/ +if (unlikely ((arg3 >XEN_ARGO_MAXIOV ) || (arg4 & ~0xffffffffUL )) )
I still think that using either UINT32_MAX, GB(4) or >> 32 would be better than an open-coded mask.
Roger.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 09/15] argo: implement the sendv op; evtchn: expose send_guest_global_virq
Quoted 36 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 17:35, <[email protected]> wrote:
Quoted 22 lines · Christopher Clark
xen/common/argo.c +22
-0+static int +ringbuf_insert (const struct domain *d,struct argo_ring_info *ring_info, +const struct argo_ring_id *src_id,xen_argo_iov_t *iovs, +unsigned int niov,uint32_t message_type, +unsigned long *out_len) +{ +xen_argo_ring_t ring; +struct xen_argo_ring_message_header mh = { }; +int sp, ret; +unsigned int len =0 ; +xen_argo_iov_t *piov; +XEN_GUEST_HANDLE (uint8) NULL_hnd = { }; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* +* Obtain the total size of data to transmit -- sets the 'len' variable +* -- and sanity check that the iovs conform to size and number limits. +* Enforced below: no more than 'len' bytes of guest data +* (plus the message header) will be sent in this operation. +*/ + ret =iov_count (iovs, niov, &len);xen/common/argo.c update_tx_ptr(const struct domain *d, struct argo_ring_info *ring_info, · line 716
Quoted 10 lines
xen/common/argo.c +10
-0+if ( ret ) +return ret; + +/* +* Upper bound check the message len against the ring size. +* The message must not fill the ring; there must be at least one slot +* remaining so we can distinguish a full ring from an empty one. +* iov_count has already verified: len <= MAX_ARGO_MESSAGE_SIZE. +*/ + if ( (ROUNDUP_MESSAGE(len) + sizeof(struct xen_argo_ring_message_header))missing space
^
Quoted 4 lines
Quoted 1 line
+ >=
ring_info->len)Align of >= also looks weird, should be aligned to the parenthesis before
ROUNDUP_.
Well, to be precise the >= belongs at the end of the previous line, so perhaps the line wrapping wants to be changed altogether.
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 09/15] argo: implement the sendv op; evtchn: expose send_guest_global_virq
Quoted 47 lines · Roger Pau Monné
Quoted 45 lines · Christopher Clark
sendv operation is invoked to perform a synchronous send of buffers contained in iovs to a remote domain's registered ring.
It takes:
* A destination address (domid, port) for the ring to send to. It performs a most-specific match lookup, to allow for wildcard.
* A source address, used to inform the destination of where to reply.
* The address of an array of iovs containing the data to send
* .. and the length of that array of iovs
* and a 32-bit message type, available to communicate message context data (eg. kernel-to-kernel, separate from the application data).If insufficient space exists in the destination ring, it will return
-EAGAINand Xen will notify the caller when sufficient space becomes available.Accesses to the ring indices are appropriately atomic. The rings are mapped into Xen's private address space to write as needed and the mappings are retained for later use.
Notifications are sent to guests via VIRQ and
send_guest_global_virqis exposed in the change to enable argo to call it.VIRQ_ARGOis claimed from the VIRQ previously reserved for this purpose (#11).The VIRQ notification method is used rather than sending events using evtchn functions directly because:
* no current event channel type is an exact fit for the intended behaviour.
ECS_IPIis closest, but it disallows migration to other VCPUs which is not necessarily a requirement for Argo.* at the point of
argo_init, allocation of an event channel is complicated by none of the guest VCPUs being initialized yet and the event channel logic expects that a valid event channel has a present VCPU.* at the point of signalling a notification, the VIRQ logic is already
defensive: if d->vcpu[0] is NULL, the notification is just silently dropped, whereas theevtchn_sendlogic is not so defensive: vcpu[0] must not be NULL, otherwise a null pointer dereference occurs.Using a VIRQ removes the need for the guest to query to determine which event channel notifications will be delivered on. This is also likely to simplify establishing future L0/L1 nested hypervisor argo communication.
Quoted 6 lines
Quoted 2 lines
Signed-off-by Christopher Clark <[email protected]>Tested-by Chris Patterson <[email protected]>There's one style nit that I think can be fixed while committing:
Reviewed-by Roger Pau Monné <[email protected]>
Thanks.
Quoted 4 lines
Despite the usage of the open-coded mask below. As with previous patches this is argos code, so I'm not going to oppose, but again I think using such open coded masks is bad, and can lead to bugs in the code. It can be fixed by a follow up patch.
Have responded with a proposed fix to address this on the other thread.
Quoted 34 lines
Quoted 23 lines
xen/common/argo.c +22
-0+static int +ringbuf_insert (const struct domain *d,struct argo_ring_info *ring_info, +const struct argo_ring_id *src_id,xen_argo_iov_t *iovs, +unsigned int niov,uint32_t message_type, +unsigned long *out_len) +{ +xen_argo_ring_t ring; +struct xen_argo_ring_message_header mh = { }; +int sp, ret; +unsigned int len =0 ; +xen_argo_iov_t *piov; +XEN_GUEST_HANDLE (uint8) NULL_hnd = { }; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* +* Obtain the total size of data to transmit -- sets the 'len' variable +* -- and sanity check that the iovs conform to size and number limits. +* Enforced below: no more than 'len' bytes of guest data +* (plus the message header) will be sent in this operation. +*/ + ret =iov_count (iovs, niov, &len);xen/common/argo.c update_tx_ptr(const struct domain *d, struct argo_ring_info *ring_info, · line 716
Quoted 10 lines
xen/common/argo.c +10
-0+if ( ret ) +return ret; + +/* +* Upper bound check the message len against the ring size. +* The message must not fill the ring; there must be at least one slot +* remaining so we can distinguish a full ring from an empty one. +* iov_count has already verified: len <= MAX_ARGO_MESSAGE_SIZE. +*/ + if ( (ROUNDUP_MESSAGE(len) + sizeof(struct xen_argo_ring_message_header))missing space ^
Quoted 4 lines
Quoted 1 line
+ >=
ring_info->len)Align of >= also looks weird, should be aligned to the parenthesis before
ROUNDUP_.
ack
Quoted 21 lines
Quoted 21 lines
xen/common/argo.c +17
-0@@ -1175,6 +1766,42 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_sendv: + { +xen_argo_send_addr_t send_addr; +xen_argo_iov_t iovs[XEN_ARGO_MAXIOV ]; +unsigned int niov; + +XEN_GUEST_HANDLE_PARAM (xen_argo_send_addr_t) send_addr_hnd = +guest_handle_cast (arg1, xen_argo_send_addr_t); +XEN_GUEST_HANDLE_PARAM (xen_argo_iov_t) iovs_hnd = +guest_handle_cast (arg2, xen_argo_iov_t); +/* arg3 is niov */ +/* arg4 is message_type. Must be a 32-bit value. */ + + rc =copy_from_guest (&send_addr, send_addr_hnd,1 ) ? -EFAULT :0 ; +if ( rc ) +break ; +
Quoted 8 lines
xen/common/argo.c do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 1790
Quoted 5 lines
xen/common/argo.c +5
-0+/* +* Reject niov above maximum limit or message_types that are outside +* 32 bit range. +*/ +if (unlikely ((arg3 >XEN_ARGO_MAXIOV ) || (arg4 & ~0xffffffffUL )) )I still think that using either
UINT32_MAX, GB(4) or >> 32 would be better than an open-coded mask.
ack via proposal on other thread
Quoted 37 lines · Jan Beulich
Quoted 37 lines
Quoted 2 lines
Quoted 2 lines · On 31.01.19 at 17:35, <[email protected]> wrote:
Quoted 22 lines · Christopher Clark
xen/common/argo.c +22
-0+static int +ringbuf_insert (const struct domain *d,struct argo_ring_info *ring_info, +const struct argo_ring_id *src_id,xen_argo_iov_t *iovs, +unsigned int niov,uint32_t message_type, +unsigned long *out_len) +{ +xen_argo_ring_t ring; +struct xen_argo_ring_message_header mh = { }; +int sp, ret; +unsigned int len =0 ; +xen_argo_iov_t *piov; +XEN_GUEST_HANDLE (uint8) NULL_hnd = { }; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* +* Obtain the total size of data to transmit -- sets the 'len' variable +* -- and sanity check that the iovs conform to size and number limits. +* Enforced below: no more than 'len' bytes of guest data +* (plus the message header) will be sent in this operation. +*/ + ret =iov_count (iovs, niov, &len);xen/common/argo.c update_tx_ptr(const struct domain *d, struct argo_ring_info *ring_info, · line 716
Quoted 10 lines
xen/common/argo.c +10
-0+if ( ret ) +return ret; + +/* +* Upper bound check the message len against the ring size. +* The message must not fill the ring; there must be at least one slot +* remaining so we can distinguish a full ring from an empty one. +* iov_count has already verified: len <= MAX_ARGO_MESSAGE_SIZE. +*/ + if ( (ROUNDUP_MESSAGE(len) + sizeof(struct xen_argo_ring_message_header))missing space
^
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
+ >=
ring_info->len)Align of >= also looks weird, should be aligned to the parenthesis before
ROUNDUP_.Well, to be precise the >= belongs at the end of the previous line, so perhaps the line wrapping wants to be changed altogether.
ack, have rewritten this.
Christopher
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 09/15] argo: implement the sendv op; evtchn: expose send_guest_global_virq
In reply to Christopher ClarkQuoted 21 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 05:28, <[email protected]> wrote:
xen/common/argo.c +16
-0@@ -1237,6 +1864,54 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_sendv: + { +xen_argo_send_addr_t send_addr; +xen_argo_iov_t iovs[XEN_ARGO_MAXIOV ]; +unsigned int i; +unsigned int niov; + +XEN_GUEST_HANDLE_PARAM (xen_argo_send_addr_t) send_addr_hnd = +guest_handle_cast (arg1, xen_argo_send_addr_t); +/* arg2: iovs, arg3: niov, arg4: message_type */ + + rc =copy_from_guest (&send_addr, send_addr_hnd,1 ) ? -EFAULT :0 ; +if ( rc ) +break ; + +if (unlikely (arg3 >XEN_ARGO_MAXIOV ) )
xen/common/argo.c compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 1892
Quoted 10 lines
xen/common/argo.c +10
-0+ { + rc = -EINVAL ; +break ; + } + niov =array_index_nospec (arg3,XEN_ARGO_MAXIOV +1 ); + +/* +* Limited scope for compat_iovs array: enables a single copy_from_guest +* call and discards the array from the stack before calling sendv. +*/
What makes you think the array gets removed from the stack again before the call? The typical way of setting up stack frames for a function is to allocate the full chunk of space needed at the start of the function, and remove it before returning. Without the argo_dprintk() after the switch() there would be the potential of the sendv() carried out as a tail call, but you can't rely on that.
With the current XEN_ARGO_MAXIOV value of 8 the overall frame size is still tolerable, I would say. But I think you want to add BUILD_BUG_ON()s here and in the native handler, such that careless bumping of the value won't go unnoticed (but also see below).
Quoted 10 lines
xen/include/public/argo.h +4
-0@@ -43,6 +43,28 @@/* Fixed-width type for "argo port" number. Nothing to do with evtchns. */ typedef uint32_t xen_argo_port_t ; +/* +* XEN_ARGO_MAXIOV : maximum number of iovs accepted in a single sendv. +* Caution is required if this value is increased: this determines the size of +* an array of xen_argo_iov_t structs on the hypervisor stack, so could cause
xen/include/public/argo.h line 59
Quoted 10 lines
xen/include/public/argo.h +10
-0+* stack overflow if the value is too large. +* The Linux Argo driver never passes more than two iovs. +* +* This value should also not exceed 128 to ensure that the total amount of data +* posted in a single Argo sendv operation cannot exceed 2^31 bytes, to reduce +* risk of integer overflow defects: +* Each argo iov can hold ~ 2^24 bytes, so XEN_ARGO_MAXIOV <= 2^(31-24), +* ie. keep XEN_ARGO_MAXIOV <= 128. +*/ +#define XEN_ARGO_MAXIOV 8U
How does 2^31 come into play here? uint32_t can hold up to 2^32, and you shouldn't be using signed arithmetic anywhere by this time anymore. I'm also struggling to see what the "~ 2^24 bytes" refers to - I see nothing along these lines added to the public header, and ...
xen/include/public/argo.h line 64
Quoted 4 lines
xen/include/public/argo.h +4
-0+typedef struct xen_argo_iov +{ + XEN_GUEST_HANDLE(uint8) iov_hnd; + uint32_t iov_len;
... the field here allows for 2^32-1. Oh, it's XEN_ARGO_MAX_RING_SIZE. It would help if the comment cross referenced that name.
Btw., neither of these two maximum values look to be architectural limits, so I wonder whether, before declaring the ABI stable, these constants shouldn't be purged and replaced by settings the guest is to retrieve via hypercall.
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 09/15] argo: implement the sendv op; evtchn: expose send_guest_global_virq
Quoted 31 lines · Jan Beulich
Quoted 31 lines
Quoted 2 lines
Quoted 2 lines · On 31.01.19 at 05:28, <[email protected]> wrote:
xen/common/argo.c +25
-0@@ -1237,6 +1864,54 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_sendv: + { +xen_argo_send_addr_t send_addr; +xen_argo_iov_t iovs[XEN_ARGO_MAXIOV ]; +unsigned int i; +unsigned int niov; + +XEN_GUEST_HANDLE_PARAM (xen_argo_send_addr_t) send_addr_hnd = +guest_handle_cast (arg1, xen_argo_send_addr_t); +/* arg2: iovs, arg3: niov, arg4: message_type */ + + rc =copy_from_guest (&send_addr, send_addr_hnd,1 ) ? -EFAULT :0 ; +if ( rc ) +break ; + +if (unlikely (arg3 >XEN_ARGO_MAXIOV ) ) + { + rc = -EINVAL ; +break ; + } + niov =array_index_nospec (arg3,XEN_ARGO_MAXIOV +1 ); + +/* +* Limited scope for compat_iovs array: enables a single copy_from_guest +* call and discards the array from the stack before calling sendv.
Quoted 8 lines
Review compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 1892
Quoted 1 line
(quoted diff) +1
-0+ */What makes you think the array gets removed from the stack again before the call? The typical way of setting up stack frames for a function is to allocate the full chunk of space needed at the start of the function, and remove it before returning. Without the
argo_dprintk()after theswitch()there would be the potential of thesendv()carried out as a tail call, but you can't rely on that.
OK. I've revised the comment.
Quoted 5 lines
With the current
XEN_ARGO_MAXIOVvalue of 8 the overall frame size is still tolerable, I would say. But I think you want to add BUILD_BUG_ON()s here and in the native handler, such that careless bumping of the value won't go unnoticed (but also see below).
ack, I've added BUILD_BUG_ON to both.
Quoted 27 lines
Quoted 11 lines
xen/include/public/argo.h +4
-0@@ -43,6 +43,28 @@/* Fixed-width type for "argo port" number. Nothing to do with evtchns. */ typedef uint32_t xen_argo_port_t ; +/* +* XEN_ARGO_MAXIOV : maximum number of iovs accepted in a single sendv. +* Caution is required if this value is increased: this determines the size of +* an array of xen_argo_iov_t structs on the hypervisor stack, so could cause xen/include/public/argo.h line 59
Quoted 10 lines
xen/include/public/argo.h +10
-0+* stack overflow if the value is too large. +* The Linux Argo driver never passes more than two iovs. +* +* This value should also not exceed 128 to ensure that the total amount of data +* posted in a single Argo sendv operation cannot exceed 2^31 bytes, to reduce +* risk of integer overflow defects: +* Each argo iov can hold ~ 2^24 bytes, so XEN_ARGO_MAXIOV <= 2^(31-24), +* ie. keep XEN_ARGO_MAXIOV <= 128. +*/ +#define XEN_ARGO_MAXIOV 8UHow does 2^31 come into play here?
uint32_tcan hold up to 2^32, and you shouldn't be using signed arithmetic anywhere by this time anymore. I'm also struggling to see what the "~ 2^24 bytes" refers to - I see nothing along these lines added to the public header, and ...
Quoted 7 lines
xen/include/public/argo.h line 64
Quoted 4 lines
xen/include/public/argo.h +4
-0+typedef struct xen_argo_iov +{ + XEN_GUEST_HANDLE(uint8) iov_hnd; + uint32_t iov_len;... the field here allows for 2^32-1. Oh, it's
XEN_ARGO_MAX_RING_SIZE. It would help if the comment cross referenced that name.
I've removed the second paragraph of the comment entirely as it's no longer accurate or required due to the bounds checking in iov_count.
Quoted 4 lines
Btw., neither of these two maximum values look to be architectural limits, so I wonder whether, before declaring the ABI stable, these constants shouldn't be purged and replaced by settings the guest is to retrieve via hypercall.
That could potentially be useful; though it hasn't been necessary so far. (fwiw: A determined guest can already retrieve these settings via hypercall.)
To make Argo's current Experimental status clearer, with the ABI stability status that accords, I propose the following addition to SUPPORT.md:
Within section: ## Virtual Hardware, Hypervisor
### Argo: Inter-domain message delivery by hypercall.
Status: Experimental
Christopher
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 09/15] argo: implement the sendv op; evtchn: expose send_guest_global_virq
Quoted 9 lines
Quoted 1 line
Quoted 1 line · On 05.02.19 at 01:52, <[email protected]> wrote:
To make Argo's current Experimental status clearer, with the ABI stability status that accords, I propose the following addition to SUPPORT.md:
Within section: ## Virtual Hardware, Hypervisor
### Argo: Inter-domain message delivery by hypercall.
Status: Experimental
Ah yes. It's odd enough that no-one so far has noticed the need for an addition there. I think we're still on the path of learning that feature additions need new entries there.
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 10/15 argo: implement the notify op
In reply to Christopher ClarkQueries for data about space availability in registered rings and causes notification to be sent when space has become available.
The hypercall op populates a supplied data structure with information about ring state and if insufficient space is currently available in a given ring, the hypervisor will record the domain's expressed interest and notify it when it observes that space has become available.
Checks for free space occur when this notify op is invoked, so it may be intentionally invoked with no data structure to populate
(ie. a NULL argument) to trigger such a check and consequent notifications.
Limit the maximum number of notify requests in a single operation to a simple fixed limit of 256.
---
v6 #09 Jan: add compat ABI
v6 #04 Jan: xlat.lst: move argo struct entries to alphabetical position
v6: rewrap comment to fix line length
v6 #10 Roger: use list_for_each_entry{_safe} for looping
v5: add EBUSY ent flag when too many domains are already on pending list
v5: reorder notify flags: error flags last, fixed state first
v5: add compat validation macros to primary source file: common/argo.c v5 : convert hypercall arg structs to struct form for compat checking
v5: dropped external file for compat macros: common/compat/argo.c v4 #10 Roger: consolidate notify flags; infer pending notify if needed v4 bugfix: take L3 before accessing ring_info in fill_ring_data
v4 #10 Roger: shorten notify flag names: drop _DATA_F
v4 #10 self/Roger: fill_ring_data: check pending_requeue error code v4 : use standard data structures as per common code
v4 #10 Roger: lower indentation in fill_ring_data by using goto
v4 #10 Roger: reword the XEN_ARGO_RING_DATA_F_SUFFICIENT comment v4 fix location of a FIXME that was incorrectly moved by this later commit v3 #07 Jan: fix format string indention in printks
v3 (general) Jan: drop fixed width types for ringbuf_payload_space v3 #07 Jan: rename ring_find_info_by_match to find_ring_info_by_match v3 #07 Jan: fix numeric entries in printk format strings
v3: ringbuf_payload_space: simpler return 0 if get_sanitized_ring fails v3 #10 Roger: simplify ringbuf_payload_space for empty rings
v3 #10 Roger: ringbuf_payload_space: add comment to explain how ret < INT32_MAX v3 #10 Roger: drop out label, use return -EFAULT in fill_ring_data v3 #10 Roger: add newline in signal_domid
v3 #10 Roger: move find functions to top of file and drop prototypes v3 #04 Jan: meld the compat hypercall arg checking
v3 #04 Roger/Jan: make lock names clearer and assert their state v3 #04 Jan: port -> aport with type; distinguish argo port from evtchn v3 self: drop braces in foreach of notify_check_pending
v3 feedback Roger/Jan: ASSERT currd is current->domain or use 'd' variable name v2 feedback Jan: drop cookie, implement teardown
v2 notify: add flag to indicate ring is shared
v2 argument name for fill_ring_data arg is now currd
v2 self: check ring size vs request and flag error rather than queue signal v2 feedback Jan: drop 'message' from 'argo_message_op'
v2 self: simplify signal_domid, drop unnecessary label + goto
v2 self: skip the cookie check in pending_cancel
v2 self: implement npending limit on number of pending entries
v1 feedback #16 Jan: sanitize_ring in ringbuf_payload_space
v2 self: inline fill_ring_data_array
v2 self: avoid retesting dst_d for put_domain
v2 self/Jan: remove use of magic verification field and tidy up
v1 feedback #16 Jan: remove testing of magic in guest-supplied structure v2 self: s/argo_pending_ent/pending_ent/g
v2 feedback v1#13 Roger: use OS-supplied roundup; drop from public header v1,2 feedback Jan/Roger/Paul: drop errno returning guest access functions v1 feedback Roger, Jan: drop argo prefix on static functions
v2 self: reduce indentation via goto out if arg NULL
v1 feedback #13 Jan: resolve checking of array handle and use of __copy v1 #5 (#16) feedback Paul: notify op: use currd in do_argo_message_op v1 #5 (#16) feedback Paul: notify op: use currd in argo_notify
v1 #5 (#16) feedback Paul: notify op: use currd in argo_notify_check_pending v1 #5 (#16) feedback Paul: notify op: use currd in argo_fill_ring_data_array v1 #13 (#16) feedback Paul: notify op: do/while: reindent only
v1 #13 (#16) feedback Paul: notify op: do/while: goto
v1 : add compat xlat.lst entries
v1: add definition for copy_field_from_guest_errno
v1 #13 feedback Jan: make 'ring data' comment comply with single-line style v1 feedback #13 Jan: use __copy; so define and use __copy_field_to_guest_errno
v1: #13 feedback Jan: public namespace: prefix with xen
v1: #13 feedback Jan: add blank line after case in do_argo_message_op
v1: self: rename ent id to domain_id
v1: self: ent id-> domain_id
v1: self: drop signal if domain_cookie mismatches
v1. feedback #15 Jan: make loop i unsigned
v1. self: drop unnecessary mb() in argo_notify_check_pending
v1. self: add blank line
v1 #16 feedback Jan: const domain arg to +argo_fill_ring_data
v1. feedback #15 Jan: check unusued hypercall args are zero
v1 feedback #16 Jan: add comment on space available signal policy v1. feedback #16 Jan: move declr, drop braces, lower indent
v1. feedback #18 Jan: meld the resource limits into the main commit v1. feedback #16 Jan: clarify use of magic field
v1. self: use single copy to read notify ring data struct
v1: argo_fill_ring_data: fix dprintk types for port field
v1: self: use %x for printing port as per other print sites
v1. feedback Jan: add comments explaining ring full vs empty
v1. following Jan: fix argo_ringbuf_payload_space calculation for empty ring
xen/common/argo.c | 385 ++++++++++++++++++++++++++++++++++++++++++++++ xen/include/public/argo.h | 67 ++++++++ xen/include/xlat.lst | 2 + 3 files changed, 454 insertions(+)
xen/common/argo.c +385 -0
@@ -36,12 +36,17 @@ CHECK_argo_addr;#define CHECK_argo_addr struct xen_argo_addr CHECK_argo_register_ring; CHECK_argo_ring; +CHECK_argo_ring_data_ent; +#undef CHECK_argo_ring_data_ent +#define CHECK_argo_ring_data_ent struct xen_argo_ring_data_ent +CHECK_argo_ring_data; CHECK_argo_ring_message_header; CHECK_argo_unregister_ring; CHECK_argo_send_addr;#endif #define MAX_RINGS_PER_DOMAIN 128U +#define MAX_NOTIFY_COUNT 256U#define MAX_PENDING_PER_RING 32U/* All messages on the ring are padded to a multiple of the slot size. */
@@ -60,6 +65,8 @@ DEFINE_XEN_GUEST_HANDLE(xen_argo_addr_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_iov_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_register_ring_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_ring_t); +DEFINE_XEN_GUEST_HANDLE (xen_argo_ring_data_t); +DEFINE_XEN_GUEST_HANDLE (xen_argo_ring_data_ent_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_send_addr_t);DEFINE_XEN_GUEST_HANDLE (xen_argo_unregister_ring_t);#ifdef CONFIG_COMPAT
@@ -434,6 +441,18 @@ signal_domain(struct domain *d) }static void +signal_domid (domid_t domain_id) +{ +struct domain *d =get_domain_by_id (domain_id); + +if ( !d ) +return ; + +signal_domain (d); +put_domain (d); +} + +static void ring_unmap (const struct domain *d,struct argo_ring_info *ring_info) {unsigned int i;
@@ -633,6 +652,66 @@ get_sanitized_ring(const struct domain *d, xen_argo_ring_t *ring,return 0 ; } +static unsigned int +ringbuf_payload_space (const struct domain *d,struct argo_ring_info *ring_info) +{ +xen_argo_ring_t ring; +unsigned int len; +int ret; + +ASSERT (LOCKING_L3 (d, ring_info)); + + len = ring_info->len; +if ( !len ) +return 0 ; + +if (get_sanitized_ring (d, &ring, ring_info) ) +return 0 ; + +argo_dprintk ("sanitized ringbuf_payload_space: tx_ptr=%u rx_ptr=%u\n" , + ring.tx_ptr, ring.rx_ptr); + +/* +* rx_ptr == tx_ptr means that the ring has been emptied. +* See message size checking logic in the entry to ringbuf_insert which +* ensures that there is always one message slot of size ROUNDUP_MESSAGE(1) +* left available, preventing a ring from being entirely filled. +* This ensures that matching ring indexes always indicate an empty ring +* and never a full one. +*/ + ret = ring.rx_ptr - ring.tx_ptr; +if ( ret <=0 ) + ret += len; + +/* +* In a sanitized ring, we can rely on: +* (rx_ptr < ring_info->len) && +* (tx_ptr < ring_info->len) && +* (ring_info->len <= XEN_ARGO_MAX_RING_SIZE) +* +* and since: XEN_ARGO_MAX_RING_SIZE < INT32_MAX +* therefore right here: ret < INT32_MAX +* and we are safe to return it as a unsigned value from this function. +* The subtractions below cannot increase its value. +*/ + +/* +* The maximum size payload for a message that will be accepted is: +* (the available space between the ring indexes) +* minus (space for a message header) +* minus (space for one message slot) +* since ringbuf_insert requires that one message slot be left +* unfilled, to avoid filling the ring to capacity and confusing a full +* ring with an empty one. +* Since the ring indexes are sanitized, the value in ret is aligned, so +* the simple subtraction here works to return the aligned value needed: +*/ + ret -=sizeof (struct xen_argo_ring_message_header ); + ret -=ROUNDUP_MESSAGE (1 ); + +return (ret <0 ) ?0 : ret; +} +/* * iov_count returns its count on success via an out variable to avoid * potential for a negative return value to be used incorrectly
@@ -965,6 +1044,63 @@ pending_remove_all(const struct domain *d, struct argo_ring_info *ring_info) ring_info->npending =0 ; } +static void +pending_notify (struct list_head *to_notify) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_Read_L1); + +/* Sending signals for all ents in this list, draining until it is empty. */ +list_for_each_entry_safe (ent, next, to_notify, node) + { +list_del (&ent->node); +signal_domid (ent->domain_id); +xfree (ent); + } +} + +static void +pending_find (const struct domain *d,struct argo_ring_info *ring_info, +unsigned int payload_space,struct list_head *to_notify) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_Read_rings_L2 (d)); + +/* +* TODO: Current policy here is to signal _all_ of the waiting domains +* interested in sending a message of size less than payload_space. +* +* This is likely to be suboptimal, since once one of them has added +* their message to the ring, there may well be insufficient room +* available for any of the others to transmit, meaning that they were +* woken in vain, which created extra work just to requeue their wait. +* +* Retain this simple policy for now since it at least avoids starving a +* domain of available space notifications because of a policy that only +* notified other domains instead. Improvement may be possible; +* investigation required. +*/ +spin_lock (&ring_info->L3_lock); + +/* Remove matching ents from the ring list, and add them to "to_notify" */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node) + { +if ( payload_space >= ent->len ) + { +if ( ring_info->id.partner_id ==XEN_ARGO_DOMID_ANY ) +wildcard_pending_list_remove (ent->domain_id, ent); + +list_del (&ent->node); + ring_info->npending--; +list_add (&ent->node, to_notify); + } + } + +spin_unlock (&ring_info->L3_lock); +} +static int pending_queue (const struct domain *d,struct argo_ring_info *ring_info,domid_t src_id,unsigned int len)
@@ -1026,6 +1162,29 @@ pending_requeue(const struct domain *d, struct argo_ring_info *ring_info, }static void +pending_cancel (const struct domain *d,struct argo_ring_info *ring_info, +domid_t src_id) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_L3 (d, ring_info)); + +/* Remove all ents where domain_id matches src_id from the ring's list. */ +list_for_each_entry_safe (ent, next, &ring_info->pending, node) + { +if ( ent->domain_id == src_id ) + { +/* For wildcard rings, remove each from their wildcard list too. */ +if ( ring_info->id.partner_id ==XEN_ARGO_DOMID_ANY ) +wildcard_pending_list_remove (ent->domain_id, ent); +list_del (&ent->node); +xfree (ent); + ring_info->npending--; + } + } +} + +static void wildcard_rings_pending_remove (struct domain *d) {struct pending_ent *ent, *next;
@@ -1152,6 +1311,99 @@ partner_rings_remove(struct domain *src_d) }static int +fill_ring_data (const struct domain *currd, +XEN_GUEST_HANDLE (xen_argo_ring_data_ent_t) data_ent_hnd) +{ +xen_argo_ring_data_ent_t ent; +struct domain *dst_d; +struct argo_ring_info *ring_info; +int ret =0 ; + +ASSERT (currd == current->domain); +ASSERT (LOCKING_Read_L1); + +if (__copy_from_guest (&ent, data_ent_hnd,1 ) ) +return -EFAULT ; + +argo_dprintk ("fill_ring_data: ent.ring.domain=%u,ent.ring.aport=%x\n" , + ent.ring.domain_id, ent.ring.aport); + + ent.flags =0 ; + + dst_d =get_domain_by_id (ent.ring.domain_id); +if ( !dst_d || !dst_d->argo ) + goto out; + +read_lock (&dst_d->argo->rings_L2_rwlock); + + ring_info =find_ring_info_by_match (dst_d, ent.ring.aport, + currd->domain_id); +if ( ring_info ) + { +unsigned int space_avail; + + ent.flags |=XEN_ARGO_RING_EXISTS ; + +spin_lock (&ring_info->L3_lock); + + ent.max_message_size = ring_info->len - +sizeof (struct xen_argo_ring_message_header ) - +ROUNDUP_MESSAGE (1 ); + +if ( ring_info->id.partner_id ==XEN_ARGO_DOMID_ANY ) + ent.flags |=XEN_ARGO_RING_SHARED ; + + space_avail =ringbuf_payload_space (dst_d, ring_info); + +argo_dprintk ("fill_ring_data: aport=%x space_avail=%u" +" space_wanted=%u\n" , + ring_info->id.aport, space_avail, ent.space_required); + +/* Do not queue a notification for an unachievable size */ +if ( ent.space_required > ent.max_message_size ) + ent.flags |=XEN_ARGO_RING_EMSGSIZE ; +else if ( space_avail >= ent.space_required ) + { +pending_cancel (dst_d, ring_info, currd->domain_id); + ent.flags |=XEN_ARGO_RING_SUFFICIENT ; + } +else + { + ret =pending_requeue (dst_d, ring_info, currd->domain_id, + ent.space_required); +if ( ret == -EBUSY ) + { +/* +* Too many other domains are already awaiting notification +* about available space on this ring. Indicate this state via +* flag. No need to return an error to the caller; allow the +* processing of queries about other rings to continue. +*/ + ent.flags |=XEN_ARGO_RING_EBUSY ; + ret =0 ; + } + } + +spin_unlock (&ring_info->L3_lock); + +if ( space_avail == ent.max_message_size ) + ent.flags |=XEN_ARGO_RING_EMPTY ; + + } +read_unlock (&dst_d->argo->rings_L2_rwlock); + + out: +if ( dst_d ) +put_domain (dst_d); + +if ( !ret && (__copy_field_to_guest (data_ent_hnd, &ent, flags) || +__copy_field_to_guest (data_ent_hnd, &ent, max_message_size)) ) +return -EFAULT ; + +return ret; +} + +static int find_ring_mfn (struct domain *d,gfn_t gfn,mfn_t *mfn) {struct page_info *page;
@@ -1618,6 +1870,109 @@ register_ring(struct domain *currd,return ret; } +static void +notify_ring (const struct domain *d,struct argo_ring_info *ring_info, +struct list_head *to_notify) +{ +unsigned int space; + +ASSERT (LOCKING_Read_rings_L2 (d)); + +spin_lock (&ring_info->L3_lock); + +if ( ring_info->len ) + space =ringbuf_payload_space (d, ring_info); +else + space =0 ; + +spin_unlock (&ring_info->L3_lock); + +if ( space ) +pending_find (d, ring_info, space, to_notify); +} + +static void +notify_check_pending (struct domain *d) +{ +unsigned int i; +LIST_HEAD (to_notify); + +ASSERT (LOCKING_Read_L1); + +read_lock (&d->argo->rings_L2_rwlock); + +/* Walk all rings, call notify_ring on each to populate to_notify list */ +for ( i =0 ; i <ARGO_HASHTABLE_SIZE ; i++ ) + { +struct argo_ring_info *ring_info, *next; +struct list_head *bucket = &d->argo->ring_hash[i]; + +list_for_each_entry_safe (ring_info, next, bucket, node) +notify_ring (d, ring_info, &to_notify); + } + +read_unlock (&d->argo->rings_L2_rwlock); + +if ( !list_empty (&to_notify) ) +pending_notify (&to_notify); +} + +static long +notify (struct domain *currd, +XEN_GUEST_HANDLE_PARAM (xen_argo_ring_data_t) ring_data_hnd) +{ +XEN_GUEST_HANDLE (xen_argo_ring_data_ent_t) ent_hnd; +xen_argo_ring_data_t ring_data; +int ret =0 ; + +ASSERT (currd == current->domain); + +read_lock (&L1_global_argo_rwlock); + +if ( !currd->argo ) + { +argo_dprintk ("!d->argo, ENODEV\n" ); + ret = -ENODEV ; + goto out; + } + +notify_check_pending (currd); + +if (guest_handle_is_null (ring_data_hnd) ) + goto out; + + ret =copy_from_guest (&ring_data, ring_data_hnd,1 ) ? -EFAULT :0 ; +if ( ret ) + goto out; + +if ( ring_data.nent >MAX_NOTIFY_COUNT ) + { +gprintk (XENLOG_ERR ,"argo: notify entry count(%u) exceeds max(%u)\n" , + ring_data.nent,MAX_NOTIFY_COUNT ); + ret = -EACCES ; + goto out; + } + + ent_hnd =guest_handle_for_field (ring_data_hnd, + xen_argo_ring_data_ent_t, data[0 ]); +if (unlikely (!guest_handle_okay (ent_hnd, ring_data.nent)) ) + { + ret = -EFAULT ; + goto out; + } + +while ( !ret && ring_data.nent-- ) + { + ret =fill_ring_data (currd, ent_hnd); +guest_handle_add_offset (ent_hnd,1 ); + } + + out: +read_unlock (&L1_global_argo_rwlock); + +return ret; +} + static long sendv(struct domain *src_d, xen_argo_addr_t *src_addr, const xen_argo_addr_t *dst_addr, xen_argo_iov_t *iovs, unsigned int niov,
@@ -1802,6 +2157,21 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_notify: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_ring_data_t) ring_data_hnd = +guest_handle_cast (arg1, xen_argo_ring_data_t); + +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =notify (currd, ring_data_hnd); +break ; + } +default : rc = -EOPNOTSUPP ;break ;
@@ -1912,6 +2282,21 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_notify: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_ring_data_t) ring_data_hnd = +guest_handle_cast (arg1, xen_argo_ring_data_t); + +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =notify (currd, ring_data_hnd); +break ; + } +default : rc = -EOPNOTSUPP ;break ;
xen/include/public/argo.h +67 -0
77 lines not on this page · raw message
xen/include/xlat.lst +2 -0
@@ -34,6 +34,8 @@ ! argo_iov argo.h ? argo_register_ring argo.h ? argo_ring argo.h +? argo_ring_data argo.h +? argo_ring_data_ent argo.h ? argo_ring_message_header argo.h ? argo_send_addr argo.h ? argo_unregister_ring argo.h
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 10/15] argo: implement the notify op
Quoted 12 lines · Christopher Clark
Queries for data about space availability in registered rings and causes notification to be sent when space has become available.
The hypercall op populates a supplied data structure with information about ring state and if insufficient space is currently available in a given ring, the hypervisor will record the domain's expressed interest and notify it when it observes that space has become available.
Checks for free space occur when this notify op is invoked, so it may be intentionally invoked with no data structure to populate
(ie. a NULL argument) to trigger such a check and consequent notifications.
Quoted 5 lines
Limit the maximum number of notify requests in a single operation to a simple fixed limit of 256.
Signed-off-by Christopher Clark <[email protected]>Tested-by Chris Patterson <[email protected]>
Despite the usage of list_for_each_entry_safe instead of list_first_entry_or_null.
xen/common/argo.c pending_remove_all(const struct domain *d, struct argo_ring_info *ring_info) · line 1055
Quoted 9 lines
xen/common/argo.c +9
-0+static void +pending_notify (struct list_head *to_notify) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_Read_L1); + +/* Sending signals for all ents in this list, draining until it is empty. */ +list_for_each_entry_safe (ent, next, to_notify, node)
list_first_entry_or_null would be more suitable here.
Roger.
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 10/15] argo: implement the notify op
Quoted 14 lines · Roger Pau Monné
Quoted 12 lines · Christopher Clark
Queries for data about space availability in registered rings and causes notification to be sent when space has become available.
The hypercall op populates a supplied data structure with information about ring state and if insufficient space is currently available in a given ring, the hypervisor will record the domain's expressed interest and notify it when it observes that space has become available.
Checks for free space occur when this notify op is invoked, so it may be intentionally invoked with no data structure to populate
(ie. a NULL argument) to trigger such a check and consequent notifications.
Quoted 7 lines
Quoted 5 lines
Limit the maximum number of notify requests in a single operation to a simple fixed limit of 256.
Signed-off-by Christopher Clark <[email protected]>Tested-by Chris Patterson <[email protected]>Reviewed-by Roger Pau Monné <[email protected]>
Thanks.
Quoted 4 lines
Despite the usage of
list_for_each_entry_safeinstead oflist_first_entry_or_null.
Quoted 11 lines
xen/common/argo.c pending_remove_all(const struct domain *d, struct argo_ring_info *ring_info) · line 1055
Quoted 9 lines
xen/common/argo.c +9
-0+static void +pending_notify (struct list_head *to_notify) +{ +struct pending_ent *ent, *next; + +ASSERT (LOCKING_Read_L1); + +/* Sending signals for all ents in this list, draining until it is empty. */ +list_for_each_entry_safe (ent, next, to_notify, node)
list_first_entry_or_nullwould be more suitable here.
ack, applied.
Christopher
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 10/15] argo: implement the notify op
In reply to Christopher ClarkQuoted 31 lines
Quoted 1 line
Quoted 1 line · On 31.01.19 at 05:28, <[email protected]> wrote:
xen/common/argo.c +19
-0@@ -1802,6 +2157,21 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_notify: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_ring_data_t) ring_data_hnd = +guest_handle_cast (arg1, xen_argo_ring_data_t); + +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =notify (currd, ring_data_hnd); +break ; + } +default : rc = -EOPNOTSUPP ;break ;@@ -1912,6 +2282,21 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_notify: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_ring_data_t) ring_data_hnd = +guest_handle_cast (arg1, xen_argo_ring_data_t);
xen/common/argo.c compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 2298
Quoted 10 lines
xen/common/argo.c +10
-0+ +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =notify (currd, ring_data_hnd); +break ; + }
At the example of this (likely applies to earlier patches as much): Aren't you afraid of this recurring duplication? It's quite easy, especially when the functions here grow a little further, for someone to forget updating one (more likely the compat one obviously). Did you consider forwarding all operations not needing translation straight into do_argo_op(), and handling only the sendv one here?
Jan
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 10/15] argo: implement the notify op
Quoted 41 lines · Jan Beulich
Quoted 41 lines
Quoted 2 lines
Quoted 2 lines · On 31.01.19 at 05:28, <[email protected]> wrote:
xen/common/argo.c +28
-0@@ -1802,6 +2157,21 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_notify: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_ring_data_t) ring_data_hnd = +guest_handle_cast (arg1, xen_argo_ring_data_t); + +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =notify (currd, ring_data_hnd); +break ; + } +default : rc = -EOPNOTSUPP ;break ;@@ -1912,6 +2282,21 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,break ; } +case XEN_ARGO_OP_notify: + { +XEN_GUEST_HANDLE_PARAM (xen_argo_ring_data_t) ring_data_hnd = +guest_handle_cast (arg1, xen_argo_ring_data_t); + +if (unlikely ((!guest_handle_is_null (arg2)) || arg3 || arg4) ) + { + rc = -EINVAL ; +break ; + } + + rc =notify (currd, ring_data_hnd); +break ;
Quoted 8 lines
Review compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1, · line 2298
Quoted 1 line
(quoted diff) +1
-0+ }At the example of this (likely applies to earlier patches as much): Aren't you afraid of this recurring duplication? It's quite easy, especially when the functions here grow a little further, for someone to forget updating one (more likely the compat one obviously). Did you consider forwarding all operations not needing translation straight into
do_argo_op(), and handling only the sendv one here?
Ack, ok.
Christopher
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 11/15 xsm, argo: XSM control for argo register
In reply to Christopher ClarkXSM controls for argo ring registration with two distinct cases, where the ring being registered is:
1) Single source: registering a ring for communication to receive messages from a specified single other domain. Default policy: allow.
2) Any source: registering a ring for communication to receive messages
from any, or all, other domains (ie. wildcard).
Default policy: deny, with runtime policy configuration via bootparam.
This commit modifies the signature of core XSM hook functions in order to apply 'const' to arguments, needed in order for 'const' to be accepted in signature of functions that invoke them.
---
v6 Chris: apply const to avc_audit_data sdom and tdom struct members v6 Chris: apply const to args in dummy.h function signatures
v6 Chris: fix missing return type in xsm.h inline functions
v3 Daniel/Jan: add to the default xsm policy for the register op v3 hoist opt_argo_mac_permissive check to allow default policy to match non-XSM v3 was: Acked-by: Daniel De Graaf <[email protected]>
v3 Add Daniel's Acked-by ; note minor changes required for v4
v3 feedback #07 Roger: use opt_argo_mac_permissive : a boolean opt v2 feedback #9 Jan: refactor to use argo-mac bootparam at point of introduction v1 feedback Paul: replace use of strncmp with strcmp
v1 feedback #16 Jan: apply const to function signatures
v1 feedback #14 Jan: add blank line before return in parse_argo_mac_param
tools/flask/policy/modules/guest_features.te | 6 ++++++ xen/common/argo.c | 11 +++++++++-- xen/include/xsm/dummy.h | 14 ++++++++++++++ xen/include/xsm/xsm.h | 19 +++++++++++++++++++ xen/xsm/dummy.c | 4 ++++ xen/xsm/flask/hooks.c | 27 ++++++++++++++++++++++++--- xen/xsm/flask/include/avc.h | 4 ++-- xen/xsm/flask/policy/access_vectors | 11 +++++++++++ xen/xsm/flask/policy/security_classes | 1 + 9 files changed, 90 insertions(+), 7 deletions(-)
tools/flask/policy/modules/guest_features.te +6 -0
@@ -5,6 +5,12 @@ allow domain_type xen_t:xen tmem_op; # pmu_ctrl is for) allow domain_type xen_t:xen2 pmu_use; +# Allow all domains: +# to register single-sender (unicast) rings to partner with any domain; and +# to register any-sender (wildcard) rings that can be sent to by any domain. +allow domain_type xen_t:argo { register_any_source }; +allow domain_type domain_type:argo { register_single_source }; + # Allow guest console output to the serial console. This is used by PV Linux # and stub domains for early boot output, so don't audit even when we deny it. # Without XSM, this is enabled only if the Xen was compiled in debug mode.
xen/common/argo.c +9 -2
@@ -26,6 +26,7 @@#include <xen/nospec.h> #include <xen/sched.h> #include <xen/time.h> +#include <xsm/xsm.h> #include <public/argo.h>
@@ -1676,8 +1677,10 @@ register_ring(struct domain *currd,if ( reg.partner_id ==XEN_ARGO_DOMID_ANY ) { -if ( !opt_argo_mac_permissive ) -return -EPERM ; + ret = opt_argo_mac_permissive ?xsm_argo_register_any_source (currd) : + -EPERM ; +if ( ret ) +return ret; }else {
@@ -1688,6 +1691,10 @@ register_ring(struct domain *currd,return -ESRCH ; } + ret =xsm_argo_register_single_source (currd, dst_d); +if ( ret ) + goto out; + send_info = xzalloc(struct argo_send_info);if ( !send_info ) {
xen/include/xsm/dummy.h +14 -0
@@ -720,6 +720,20 @@ static XSM_INLINE int xsm_dm_op(XSM_DEFAULT_ARG struct domain *d)#endif /* CONFIG_X86 */ +#ifdef CONFIG_ARGO +static XSM_INLINE int xsm_argo_register_single_source(const struct domain *d, +const struct domain *t) +{ +return 0 ; +} + +static XSM_INLINE int xsm_argo_register_any_source(const struct domain *d) +{ +return 0 ; +} + +#endif /* CONFIG_ARGO */ +#include <public/version.h> static XSM_INLINE int xsm_xen_version (XSM_DEFAULT_ARG uint32_t op) {
xen/include/xsm/xsm.h +19 -0
@@ -181,6 +181,11 @@ struct xsm_operations {#endif int (*xen_version) (uint32_t cmd);int (*domain_resource_map) (struct domain *d); +#ifdef CONFIG_ARGO +int (*argo_register_single_source) (const struct domain *d, +const struct domain *t); +int (*argo_register_any_source) (const struct domain *d); +#endif };#ifdef CONFIG_XSM
@@ -698,6 +703,20 @@ static inline int xsm_domain_resource_map(xsm_default_t def, struct domain *d) return xsm_ops->domain_resource_map(d); } +#ifdef CONFIG_ARGO +static inline int xsm_argo_register_single_source (const struct domain *d, +const struct domain *t) +{ +return xsm_ops->argo_register_single_source (d, t); +} + +static inline int xsm_argo_register_any_source (const struct domain *d) +{ +return xsm_ops->argo_register_any_source (d); +} + +#endif /* CONFIG_ARGO */ +#endif /* XSM_NO_WRAPPERS */ #ifdef CONFIG_MULTIBOOT
xen/xsm/dummy.c +4 -0
@@ -152,4 +152,8 @@ void __init xsm_fixup_ops (struct xsm_operations *ops)#endif set_to_dummy_if_null (ops, xen_version);set_to_dummy_if_null (ops, domain_resource_map); +#ifdef CONFIG_ARGO +set_to_dummy_if_null (ops, argo_register_single_source); +set_to_dummy_if_null (ops, argo_register_any_source); +#endif }
xen/xsm/flask/hooks.c +24 -3
@@ -36,13 +36,14 @@#include <objsec.h> #include <conditional.h> -static u32 domain_sid (struct domain *dom) +static u32 domain_sid (const struct domain *dom) {struct domain_security_struct *dsec = dom->ssid;return dsec->sid; } -static u32 domain_target_sid (struct domain *src,struct domain *dst) +static u32 domain_target_sid (const struct domain *src, +const struct domain *dst) {struct domain_security_struct *ssec = src->ssid;struct domain_security_struct *dsec = dst->ssid;
@@ -58,7 +59,8 @@ static u32 evtchn_sid(const struct evtchn *chn)return chn->ssid.flask_sid; } -static int domain_has_perm (struct domain *dom1,struct domain *dom2, +static int domain_has_perm (const struct domain *dom1, +const struct domain *dom2,u16 class,u32 perms) {u32 ssid, tsid;
@@ -1717,6 +1719,21 @@ static int flask_domain_resource_map(struct domain *d)return current_has_perm (d,SECCLASS_DOMAIN2 ,DOMAIN2__RESOURCE_MAP ); } +#ifdef CONFIG_ARGO +static int flask_argo_register_single_source (const struct domain *d, +const struct domain *t) +{ +return domain_has_perm (d, t,SECCLASS_ARGO , +ARGO__REGISTER_SINGLE_SOURCE ); +} + +static int flask_argo_register_any_source (const struct domain *d) +{ +return avc_has_perm (domain_sid (d),SECINITSID_XEN ,SECCLASS_ARGO , +ARGO__REGISTER_ANY_SOURCE ,NULL ); +} +#endif +long do_flask_op (XEN_GUEST_HANDLE_PARAM (xsm_op_t) u_flask_op);int compat_flask_op (XEN_GUEST_HANDLE_PARAM (xsm_op_t) u_flask_op);
@@ -1851,6 +1868,10 @@ static struct xsm_operations flask_ops = {#endif .xen_version = flask_xen_version, .domain_resource_map = flask_domain_resource_map, +#ifdef CONFIG_ARGO + .argo_register_single_source = flask_argo_register_single_source, + .argo_register_any_source = flask_argo_register_any_source, +#endif };void __init flask_init (const void *policy_buffer,size_t policy_size)
xen/xsm/flask/include/avc.h +2 -2
@@ -40,8 +40,8 @@ struct avc_audit_data {#define AVC_AUDIT_DATA_RANGE 3#define AVC_AUDIT_DATA_MEMORY 4#define AVC_AUDIT_DATA_DTDEV 5 -struct domain *sdom; -struct domain *tdom; +const struct domain *sdom; +const struct domain *tdom;union {unsigned long device;int irq;
xen/xsm/flask/policy/access_vectors +11 -0
@@ -531,3 +531,14 @@ class version # Xen build id xen_build_id } + +# Class argo is used to describe the Argo interdomain communication system. +class argo +{ + # Domain requesting registration of a communication ring + # to receive messages from a specific other domain. + register_single_source + # Domain requesting registration of a communication ring + # to receive messages from any other domain. + register_any_source +}
xen/xsm/flask/policy/security_classes +1 -0
@@ -19,5 +19,6 @@ class event class grant class security class version +class argo # FLASK
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 12/15 xsm, argo: XSM control for argo message send operation
In reply to Christopher ClarkDefault policy: allow.
---
v3 Daniel/Jan: add to the default xsm policy for the send op v3 Add Daniel's Acked-by
v2: reordered commit sequence to after sendv implementation v1 feedback Jan #16: apply const to function signatures v1 version was: Reviewed-by: Paul Durrant <[email protected]>
tools/flask/policy/modules/guest_features.te | 7 ++++--- xen/common/argo.c | 11 +++++++++++ xen/include/xsm/dummy.h | 6 ++++++ xen/include/xsm/xsm.h | 6 ++++++ xen/xsm/dummy.c | 1 + xen/xsm/flask/hooks.c | 7 +++++++ xen/xsm/flask/policy/access_vectors | 2 ++ 7 files changed, 37 insertions(+), 3 deletions(-)
tools/flask/policy/modules/guest_features.te +4 -3
@@ -6,10 +6,11 @@ allow domain_type xen_t:xen tmem_op; allow domain_type xen_t:xen2 pmu_use; # Allow all domains: -# to register single-sender (unicast) rings to partner with any domain; and -# to register any-sender (wildcard) rings that can be sent to by any domain. +# to register single-sender (unicast) rings to partner with any domain; +# to register any-sender (wildcard) rings that can be sent to by any domain; +# and send messages to rings. allow domain_type xen_t:argo { register_any_source }; -allow domain_type domain_type:argo { register_single_source }; +allow domain_type domain_type:argo { send register_single_source }; # Allow guest console output to the serial console. This is used by PV Linux # and stub domains for early boot output, so don't audit even when we deny it.
xen/common/argo.c +11 -0
@@ -2014,6 +2014,17 @@ sendv(struct domain *src_d, xen_argo_addr_t *src_addr,if ( !dst_d )return -ESRCH ; + ret =xsm_argo_send (src_d, dst_d); +if ( ret ) + { +gprintk (XENLOG_ERR ,"argo: XSM REJECTED %i -> %i\n" , + src_d->domain_id, dst_d->domain_id); + +put_domain (dst_d); + +return ret; + } +read_lock (&L1_global_argo_rwlock);if ( !src_d->argo )
xen/include/xsm/dummy.h +6 -0
@@ -732,6 +732,12 @@ static XSM_INLINE int xsm_argo_register_any_source(const struct domain *d)return 0 ; } +static XSM_INLINE int xsm_argo_send(const struct domain *d, +const struct domain *t) +{ +return 0 ; +} +#endif /* CONFIG_ARGO */ #include <public/version.h>
xen/include/xsm/xsm.h +6 -0
@@ -185,6 +185,7 @@ struct xsm_operations {int (*argo_register_single_source) (const struct domain *d,const struct domain *t);int (*argo_register_any_source) (const struct domain *d); +int (*argo_send) (const struct domain *d,const struct domain *t);#endif };
@@ -715,6 +716,11 @@ static inline int xsm_argo_register_any_source(const struct domain *d)return xsm_ops->argo_register_any_source (d); } +static inline int xsm_argo_send (const struct domain *d,const struct domain *t) +{ +return xsm_ops->argo_send (d, t); +} +#endif /* CONFIG_ARGO */ #endif /* XSM_NO_WRAPPERS */
xen/xsm/dummy.c +1 -0
@@ -155,5 +155,6 @@ void __init xsm_fixup_ops (struct xsm_operations *ops)#ifdef CONFIG_ARGO set_to_dummy_if_null (ops, argo_register_single_source);set_to_dummy_if_null (ops, argo_register_any_source); +set_to_dummy_if_null (ops, argo_send);#endif }
xen/xsm/flask/hooks.c +7 -0
@@ -1732,6 +1732,12 @@ static int flask_argo_register_any_source(const struct domain *d)return avc_has_perm (domain_sid (d),SECINITSID_XEN ,SECCLASS_ARGO , ARGO__REGISTER_ANY_SOURCE, NULL); } + +static int flask_argo_send (const struct domain *d,const struct domain *t) +{ +return domain_has_perm (d, t,SECCLASS_ARGO ,ARGO__SEND ); +} +#endif long do_flask_op (XEN_GUEST_HANDLE_PARAM (xsm_op_t) u_flask_op);
@@ -1871,6 +1877,7 @@ static struct xsm_operations flask_ops = {#ifdef CONFIG_ARGO .argo_register_single_source = flask_argo_register_single_source, .argo_register_any_source = flask_argo_register_any_source, + .argo_send = flask_argo_send,#endif };
xen/xsm/flask/policy/access_vectors +2 -0
@@ -541,4 +541,6 @@ class argo # Domain requesting registration of a communication ring # to receive messages from any other domain. register_any_source + # Domain sending a message to another domain. + send }
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 13/15 xsm, argo: XSM control for any access to argo by a domain
In reply to Christopher ClarkWill inhibit initialization of the domain's argo data structure to prevent receiving any messages or notifications and access to any of the argo hypercall operations.
---
v6 #09 Jan: add compat ABI
v6 Chris: apply const to args in dummy.h function signatures v6 Chris: fix missing return type in xsm.h inline functions v3 Daniel/Jan: add to the default xsm policy for enable
v3 Add Daniel's Acked-by
v3 #04 Jason/Roger: soft_reset: can assume reinit is ok if d->argo set v2 self: fix xsm use in soft-reset prior to introduction
v1 #5 (#17) feedback Paul: XSM control for any access: use currd v1 #16 feedback Jan: apply const to function signatures
tools/flask/policy/modules/guest_features.te | 4 ++-- xen/common/argo.c | 12 ++++++------ xen/include/xsm/dummy.h | 5 +++++ xen/include/xsm/xsm.h | 6 ++++++ xen/xsm/dummy.c | 1 + xen/xsm/flask/hooks.c | 7 +++++++ xen/xsm/flask/policy/access_vectors | 3 +++ 7 files changed, 30 insertions(+), 8 deletions(-)
tools/flask/policy/modules/guest_features.te +2 -2
@@ -5,11 +5,11 @@ allow domain_type xen_t:xen tmem_op; # pmu_ctrl is for) allow domain_type xen_t:xen2 pmu_use; -# Allow all domains: +# Allow all domains to enable the Argo interdomain communication hypercall; # to register single-sender (unicast) rings to partner with any domain; # to register any-sender (wildcard) rings that can be sent to by any domain; # and send messages to rings. -allow domain_type xen_t:argo { register_any_source }; +allow domain_type xen_t:argo { enable register_any_source }; allow domain_type domain_type:argo { send register_single_source }; # Allow guest console output to the serial console. This is used by PV Linux
xen/common/argo.c +6 -6
@@ -2094,7 +2094,7 @@ do_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,argo_dprintk ("->do_argo_op(%u,%p,%p,%lu,0x%lx)\n" , cmd, (void *)arg1.p, (void *)arg2.p, arg3, arg4); -if (unlikely (!opt_argo) ) +if (unlikely (!opt_argo ||xsm_argo_enable (currd)) )return -EOPNOTSUPP ; switch (cmd)
@@ -2212,7 +2212,7 @@ compat_argo_op(unsigned int cmd, XEN_GUEST_HANDLE_PARAM(void) arg1,argo_dprintk ("->compat_argo_op(%u,%p,%p,%lu,0x%lx)\n" , cmd, (void *)arg1.p, (void *)arg2.p, arg3, arg4); -if (unlikely (!opt_argo) ) +if (unlikely (!opt_argo ||xsm_argo_enable (currd)) )return -EOPNOTSUPP ; switch (cmd)
@@ -2348,7 +2348,7 @@ argo_init(struct domain *d) {struct argo_domain *argo; -if ( !opt_argo ) +if ( !opt_argo ||xsm_argo_enable (d) ) {argo_dprintk ("argo disabled, domid: %u\n" , d->domain_id);return 0 ;
@@ -2405,9 +2405,9 @@ argo_soft_reset(struct domain *d)wildcard_rings_pending_remove (d);/* -* Since opt_argo cannot change at runtime, if d->argo is true then -* opt_argo must be true, and we can assume that init is allowed to -* proceed again here. +* Since neither opt_argo or xsm_argo_enable(d) can change at runtime, +* if d->argo is true then both opt_argo and xsm_argo_enable(d) must be +* true, and we can assume that init is allowed to proceed again here. */ argo_domain_init (d->argo); }
xen/include/xsm/dummy.h +5 -0
@@ -721,6 +721,11 @@ static XSM_INLINE int xsm_dm_op(XSM_DEFAULT_ARG struct domain *d)#endif /* CONFIG_X86 */ #ifdef CONFIG_ARGO +static XSM_INLINE int xsm_argo_enable(const struct domain *d) +{ +return 0 ; +} + static XSM_INLINE int xsm_argo_register_single_source(const struct domain *d,const struct domain *t) {
xen/include/xsm/xsm.h +6 -0
@@ -182,6 +182,7 @@ struct xsm_operations {int (*xen_version) (uint32_t cmd);int (*domain_resource_map) (struct domain *d);#ifdef CONFIG_ARGO +int (*argo_enable) (const struct domain *d);int (*argo_register_single_source) (const struct domain *d,const struct domain *t);int (*argo_register_any_source) (const struct domain *d);
@@ -705,6 +706,11 @@ static inline int xsm_domain_resource_map(xsm_default_t def, struct domain *d) }#ifdef CONFIG_ARGO +static inline int xsm_argo_enable (const struct domain *d) +{ +return xsm_ops->argo_enable (d); +} +static inline int xsm_argo_register_single_source (const struct domain *d,const struct domain *t) {
xen/xsm/dummy.c +1 -0
@@ -153,6 +153,7 @@ void __init xsm_fixup_ops (struct xsm_operations *ops)set_to_dummy_if_null (ops, xen_version);set_to_dummy_if_null (ops, domain_resource_map);#ifdef CONFIG_ARGO +set_to_dummy_if_null (ops, argo_enable);set_to_dummy_if_null (ops, argo_register_single_source);set_to_dummy_if_null (ops, argo_register_any_source);set_to_dummy_if_null (ops, argo_send);
xen/xsm/flask/hooks.c +7 -0
@@ -1720,6 +1720,12 @@ static int flask_domain_resource_map(struct domain *d) }#ifdef CONFIG_ARGO +static int flask_argo_enable (const struct domain *d) +{ +return avc_has_perm (domain_sid (d),SECINITSID_XEN ,SECCLASS_ARGO , +ARGO__ENABLE ,NULL ); +} +static int flask_argo_register_single_source (const struct domain *d,const struct domain *t) {
@@ -1875,6 +1881,7 @@ static struct xsm_operations flask_ops = { .xen_version = flask_xen_version, .domain_resource_map = flask_domain_resource_map, #ifdef CONFIG_ARGO + .argo_enable = flask_argo_enable, .argo_register_single_source = flask_argo_register_single_source, .argo_register_any_source = flask_argo_register_any_source, .argo_send = flask_argo_send,
xen/xsm/flask/policy/access_vectors +3 -0
@@ -535,6 +535,9 @@ class version # Class argo is used to describe the Argo interdomain communication system. class argo { + # Enable initialization of a domain's argo subsystem and + # permission to access the argo hypercall operations. + enable # Domain requesting registration of a communication ring # to receive messages from a specific other domain. register_single_source
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 14/15 xsm, argo: notify: don't describe rings that cannot be sent to
In reply to Christopher Clark---
v3 #10 Roger: drop out label, use return -EFAULT in fill_ring_data
v3: Add Daniel's Acked-by
xen/common/argo.c | 11 +++++++++++ 1 file changed, 11 insertions(+)
xen/common/argo.c +11 -0
@@ -1335,6 +1335,17 @@ fill_ring_data(const struct domain *currd,if ( !dst_d || !dst_d->argo ) goto out; +/* +* Don't supply information about rings that a guest is not +* allowed to send to. +*/ + ret =xsm_argo_send (currd, dst_d); +if ( ret ) + { +put_domain (dst_d); +return ret; + } +read_lock (&dst_d->argo->rings_L2_rwlock); ring_info = find_ring_info_by_match(dst_d, ent.ring.aport,
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
PATCH v7 15/15 MAINTAINERS: add new section for Argo and self as maintainer
In reply to Christopher Clark---
v5 whitespace: tabs
MAINTAINERS | 7 +++++++ 1 file changed, 7 insertions(+)
MAINTAINERS +7 -0
@@ -158,6 +158,13 @@ S: Supported F: xen/arch/x86/hvm/svm/ F: xen/arch/x86/cpu/vpmu_amd.c +ARGO +M: Christopher Clark <[email protected]> +S: Maintained +F: xen/include/public/argo.h +F: xen/include/xen/argo.h +F: xen/common/argo.c + ARINC653 SCHEDULER M: Josh Whitehead <[email protected]> M: Robert VanVossen <[email protected]>
Signature · 7 lines
2.7.4 _______________________________________________ Xen-devel mailing list [email protected] https://lists.xenproject.org/mailman/listinfo/xen-devel
Re: [PATCH v7 15/15] MAINTAINERS: add new section for Argo and self as maintainer
Quoted 1 line · Christopher Clark
Signed-off-by Christopher Clark <[email protected]>
_______________________________________________
Xen-devel mailing list
[email protected]
https://lists.xenproject.org/mailman/listinfo/xen-devel