[PATCH v6 0/4] Enable guest suspend/resume support on ARM via vPSCI

[PATCH v6 0/4] Enable guest suspend/resume support on ARM via vPSCI

[Mykola Kvach]

From: Mykola Kvach <mykola_kvach@epam.com>

This patch series introduces the initial support for guest suspend
and resume on ARM platforms using the PSCI SYSTEM_SUSPEND interface. The main
goal is to allow ARM guests to request suspension using PSCI and be resumed
by the control domain (e.g., via "xl resume").

### Background

The PSCI SYSTEM_SUSPEND call is part of the PSCI v1.0+ specification and is
used by guests to enter the deepest possible power state. On Xen/ARM, we
emulate this interface in the virtual PSCI (vPSCI) layer for guests.

This series includes:

1. A new vPSCI implementation of the PSCI SYSTEM_SUSPEND function for guests
2. Documentation updates to SUPPORT.md to reflect PSCI and vPSCI support status
3. Enabling "xl resume" command compilation for ARM, which was previously disabled

### Usage

For Linux-based guests:
  - Suspend can be triggered using: "echo mem > /sys/power/state" or "systemctl suspend"
  - Resume can be performed from control domain using: "xl resume <domain>"

For more information, refer to the official Linux kernel documentation on power management.

Note that currently, SYSTEM_SUSPEND is supported only for guest domains (not for
the hardware domain).
---

TODO: enable "xl suspend" for ARM
---
Previous versions of this patch series:
  V1: https://lists.xenproject.org/archives/html/xen-devel/2018-11/msg01093.html
  V2: https://marc.info/?l=xen-devel&m=166514782207736&w=2
  V3: https://lists.xenproject.org/archives/html/xen-devel/2025-03/msg00168.html


This is the first part of previous patch series and originally consist only
with necessary changes needed for guest domain suspend.

Main changes in V6:
- Skip execution of ctxt_switch_from for VCPUs in paused domains.
- Implement domain_resume_nopause
- Add a helper to determine if a VCPU is in suspended domain.
- Ignore upper 32 bits of arguments for 64-bit domains calling SMC32 SYSTEM_SUSPEND.
- Macro renamed from LIBXL_HAVE_NO_SUSPEND_RESUME to LIBXL_HAVE_NO_SUSPEND for clarity.
- Documentation now focuses only on the SYSTEM_SUSPEND function, with improved wording and structure.
- Changelog and commit messages refined for clarity and to remove redundant explanations.

Main change in V5:
  - Reverted the logic related to suspending domains. Instead of the standby
    mode introduced in v4, domains now resume execution at the point provided
    during suspend

The rest of the minor changes are described in the changelog of each commit.

Mykola Kvach (4):
  xen/arm: Implement PSCI SYSTEM_SUSPEND call for guests
  tools/xl: Allow compilation of 'xl resume' command on Arm
  SUPPORT.md: Document PSCI SYSTEM_SUSPEND support for guests
  CHANGELOG: Document guest suspend/resume to RAM support on Arm

 CHANGELOG.md                          |  2 +
 SUPPORT.md                            |  5 +-
 tools/include/libxl.h                 |  5 +-
 tools/xl/xl.h                         | 10 +--
 tools/xl/xl_cmdtable.c                |  8 +--
 tools/xl/xl_migrate.c                 |  4 +-
 tools/xl/xl_saverestore.c             |  4 +-
 tools/xl/xl_vmcontrol.c               | 14 ++---
 xen/arch/arm/domain.c                 |  4 ++
 xen/arch/arm/include/asm/perfc_defn.h |  1 +
 xen/arch/arm/include/asm/psci.h       |  2 +
 xen/arch/arm/include/asm/vpsci.h      |  2 +-
 xen/arch/arm/vpsci.c                  | 90 +++++++++++++++++++++++----
 xen/common/domain.c                   | 30 ++++++---
 xen/include/xen/sched.h               |  2 +
 15 files changed, 137 insertions(+), 46 deletions(-)

-- 
2.48.1

[PATCH v6 1/4] xen/arm: Implement PSCI SYSTEM_SUSPEND call for guests

[Mykola Kvach]

From: Mykola Kvach <mykola_kvach@epam.com>

This patch adds support for the PSCI SYSTEM_SUSPEND function in the vPSCI
(virtual PSCI) interface, allowing guests to request suspend via the PSCI
v1.0 SYSTEM_SUSPEND call (both 32-bit and 64-bit variants).

The implementation:
- Adds SYSTEM_SUSPEND function IDs to PSCI definitions
- Implements trapping and handling of SYSTEM_SUSPEND in vPSCI
- Allows only non-hardware domains to invoke SYSTEM_SUSPEND; for the
  hardware domain, PSCI_NOT_SUPPORTED is returned to avoid halting the
  system in hwdom_shutdown() called from domain_shutdown
- Ensures all secondary VCPUs of the calling domain are offline before
  allowing suspend due to PSCI spec

Usage:

For Linux-based guests, suspend can be initiated with:
    echo mem > /sys/power/state
or via:
    systemctl suspend

Resuming the guest is performed from control domain using:
      xl resume <domain>

Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
---
Changes in V6:
- skip execution of ctxt_switch_from for vcpu that is in paused domain
- add implementation of domain_resume without domain_pause
- add helper function to determine if vcpu is suspended or not
- ignore upper 32 bits of argument values when the domain is 64-bit
  and calls the SMC32 SYSTEM_SUSPEND function
- cosmetic changes after review

Changes in V5:
- don't use standby mode, restore execution in a provided by guest point
- move checking that all CPUs, except current one, are offline to after
  pausing the vCPUs
- provide ret status from arch_domain_shutdown and handle it in
  domain_shutdown
- adjust VPSCI_NR_FUNCS to reflect the number of newly added PSCI functions

Changes in V4:
Dropped all changes related to watchdog, domain is marked as shutting
down in domain_shutdown and watchdog timeout handler won't trigger
because of it.

Previous versions included code to manage Xen watchdog timers during suspend,
but this was removed. When a guest OS starts the Xen watchdog (either via the
kernel driver or xenwatchdogd), it is responsible for managing that state
across suspend/resume. On Linux, the Xen kernel driver properly stops the
watchdog during suspend. However, when xenwatchdogd is used instead, suspend
handling is incomplete, potentially leading to watchdog-triggered resets on
resume. Xen leaves watchdog handling to the guest OS and its services.

Dropped all changes related to VCPU context, because instead domain_shutdown
is used, so we don't need any extra changes for suspending domain.

Changes in V3:
Dropped all domain flags and related code (which touched common functions like
vcpu_unblock), keeping only the necessary changes for Xen suspend/resume, i.e.
suspend/resume is now fully supported only for the hardware domain.
Proper support for domU suspend/resume will be added in a future patch.
This patch does not yet include VCPU context reset or domain context
restoration in VCPU.
---
 xen/arch/arm/domain.c                 |  4 ++
 xen/arch/arm/include/asm/perfc_defn.h |  1 +
 xen/arch/arm/include/asm/psci.h       |  2 +
 xen/arch/arm/include/asm/vpsci.h      |  2 +-
 xen/arch/arm/vpsci.c                  | 90 +++++++++++++++++++++++----
 xen/common/domain.c                   | 30 ++++++---
 xen/include/xen/sched.h               |  2 +
 7 files changed, 110 insertions(+), 21 deletions(-)

diff --git a/xen/arch/arm/domain.c b/xen/arch/arm/domain.c
index 79a144e61b..a2ad10146f 100644
--- a/xen/arch/arm/domain.c
+++ b/xen/arch/arm/domain.c
@@ -90,6 +90,10 @@ static void ctxt_switch_from(struct vcpu *p)
     if ( is_idle_vcpu(p) )
         return;
 
+    /* VCPU's context should not be saved if its domain is suspended */
+    if ( vcpu_is_suspended(p) )
+        return;
+
     p2m_save_state(p);
 
     /* CP 15 */
diff --git a/xen/arch/arm/include/asm/perfc_defn.h b/xen/arch/arm/include/asm/perfc_defn.h
index effd25b69e..8dfcac7e3b 100644
--- a/xen/arch/arm/include/asm/perfc_defn.h
+++ b/xen/arch/arm/include/asm/perfc_defn.h
@@ -33,6 +33,7 @@ PERFCOUNTER(vpsci_system_reset,        "vpsci: system_reset")
 PERFCOUNTER(vpsci_cpu_suspend,         "vpsci: cpu_suspend")
 PERFCOUNTER(vpsci_cpu_affinity_info,   "vpsci: cpu_affinity_info")
 PERFCOUNTER(vpsci_features,            "vpsci: features")
+PERFCOUNTER(vpsci_system_suspend,      "vpsci: system_suspend")
 
 PERFCOUNTER(vcpu_kick,                 "vcpu: notify other vcpu")
 
diff --git a/xen/arch/arm/include/asm/psci.h b/xen/arch/arm/include/asm/psci.h
index 4780972621..48a93e6b79 100644
--- a/xen/arch/arm/include/asm/psci.h
+++ b/xen/arch/arm/include/asm/psci.h
@@ -47,10 +47,12 @@ void call_psci_system_reset(void);
 #define PSCI_0_2_FN32_SYSTEM_OFF          PSCI_0_2_FN32(8)
 #define PSCI_0_2_FN32_SYSTEM_RESET        PSCI_0_2_FN32(9)
 #define PSCI_1_0_FN32_PSCI_FEATURES       PSCI_0_2_FN32(10)
+#define PSCI_1_0_FN32_SYSTEM_SUSPEND      PSCI_0_2_FN32(14)
 
 #define PSCI_0_2_FN64_CPU_SUSPEND         PSCI_0_2_FN64(1)
 #define PSCI_0_2_FN64_CPU_ON              PSCI_0_2_FN64(3)
 #define PSCI_0_2_FN64_AFFINITY_INFO       PSCI_0_2_FN64(4)
+#define PSCI_1_0_FN64_SYSTEM_SUSPEND      PSCI_0_2_FN64(14)
 
 /* PSCI v0.2 affinity level state returned by AFFINITY_INFO */
 #define PSCI_0_2_AFFINITY_LEVEL_ON      0
diff --git a/xen/arch/arm/include/asm/vpsci.h b/xen/arch/arm/include/asm/vpsci.h
index 0cca5e6830..69d40f9d7f 100644
--- a/xen/arch/arm/include/asm/vpsci.h
+++ b/xen/arch/arm/include/asm/vpsci.h
@@ -23,7 +23,7 @@
 #include <asm/psci.h>
 
 /* Number of function implemented by virtual PSCI (only 0.2 or later) */
-#define VPSCI_NR_FUNCS  12
+#define VPSCI_NR_FUNCS  14
 
 /* Functions handle PSCI calls from the guests */
 bool do_vpsci_0_1_call(struct cpu_user_regs *regs, uint32_t fid);
diff --git a/xen/arch/arm/vpsci.c b/xen/arch/arm/vpsci.c
index 7ba9ccd94b..11334b3426 100644
--- a/xen/arch/arm/vpsci.c
+++ b/xen/arch/arm/vpsci.c
@@ -10,28 +10,18 @@
 
 #include <public/sched.h>
 
-static int do_common_cpu_on(register_t target_cpu, register_t entry_point,
+static int do_setup_vcpu_ctx(struct vcpu *v, register_t entry_point,
                             register_t context_id)
 {
-    struct vcpu *v;
     struct domain *d = current->domain;
     struct vcpu_guest_context *ctxt;
     int rc;
     bool is_thumb = entry_point & 1;
-    register_t vcpuid;
-
-    vcpuid = vaffinity_to_vcpuid(target_cpu);
-
-    if ( (v = domain_vcpu(d, vcpuid)) == NULL )
-        return PSCI_INVALID_PARAMETERS;
 
     /* THUMB set is not allowed with 64-bit domain */
     if ( is_64bit_domain(d) && is_thumb )
         return PSCI_INVALID_ADDRESS;
 
-    if ( !test_bit(_VPF_down, &v->pause_flags) )
-        return PSCI_ALREADY_ON;
-
     if ( (ctxt = alloc_vcpu_guest_context()) == NULL )
         return PSCI_DENIED;
 
@@ -78,11 +68,32 @@ static int do_common_cpu_on(register_t target_cpu, register_t entry_point,
     if ( rc < 0 )
         return PSCI_DENIED;
 
-    vcpu_wake(v);
-
     return PSCI_SUCCESS;
 }
 
+static int do_common_cpu_on(register_t target_cpu, register_t entry_point,
+                            register_t context_id)
+{
+    int rc;
+    struct vcpu *v;
+    struct domain *d = current->domain;
+    register_t vcpuid;
+
+    vcpuid = vaffinity_to_vcpuid(target_cpu);
+
+    if ( (v = domain_vcpu(d, vcpuid)) == NULL )
+        return PSCI_INVALID_PARAMETERS;
+
+    if ( !test_bit(_VPF_down, &v->pause_flags) )
+        return PSCI_ALREADY_ON;
+
+    rc = do_setup_vcpu_ctx(v, entry_point, context_id);
+    if ( rc == PSCI_SUCCESS )
+        vcpu_wake(v);
+
+    return rc;
+}
+
 static int32_t do_psci_cpu_on(uint32_t vcpuid, register_t entry_point)
 {
     int32_t ret;
@@ -197,6 +208,39 @@ static void do_psci_0_2_system_reset(void)
     domain_shutdown(d,SHUTDOWN_reboot);
 }
 
+static int32_t do_psci_1_0_system_suspend(register_t epoint, register_t cid)
+{
+    int ret;
+    struct vcpu *v;
+    struct domain *d = current->domain;
+
+    /* SYSTEM_SUSPEND is not supported for the hardware domain yet */
+    if ( is_hardware_domain(d) )
+        return PSCI_NOT_SUPPORTED;
+
+    /*
+     * Shut down the domain first to ensure all vCPUs are paused
+     * before checking that they (except the calling one) are offline.
+     */
+    domain_shutdown(d, SHUTDOWN_suspend);
+
+    /* Ensure that all CPUs other than the calling one are offline */
+    for_each_vcpu ( d, v )
+    {
+        if ( v != current && is_vcpu_online(v) )
+        {
+            domain_resume_nopause(d);
+            return PSCI_DENIED;
+        }
+    }
+
+    ret = do_setup_vcpu_ctx(current, epoint, cid);
+    if ( ret != PSCI_SUCCESS )
+        domain_resume_nopause(d);
+
+    return ret;
+}
+
 static int32_t do_psci_1_0_features(uint32_t psci_func_id)
 {
     /* /!\ Ordered by function ID and not name */
@@ -214,6 +258,8 @@ static int32_t do_psci_1_0_features(uint32_t psci_func_id)
     case PSCI_0_2_FN32_SYSTEM_OFF:
     case PSCI_0_2_FN32_SYSTEM_RESET:
     case PSCI_1_0_FN32_PSCI_FEATURES:
+    case PSCI_1_0_FN32_SYSTEM_SUSPEND:
+    case PSCI_1_0_FN64_SYSTEM_SUSPEND:
     case ARM_SMCCC_VERSION_FID:
         return 0;
     default:
@@ -344,6 +390,24 @@ bool do_vpsci_0_2_call(struct cpu_user_regs *regs, uint32_t fid)
         return true;
     }
 
+    case PSCI_1_0_FN32_SYSTEM_SUSPEND:
+    case PSCI_1_0_FN64_SYSTEM_SUSPEND:
+    {
+        register_t epoint = PSCI_ARG(regs, 1);
+        register_t cid = PSCI_ARG(regs, 2);
+
+        if ( is_64bit_domain(current->domain) &&
+             fid == PSCI_1_0_FN32_SYSTEM_SUSPEND )
+        {
+            epoint &= GENMASK(31, 0);
+            cid &= GENMASK(31, 0);
+        }
+
+        perfc_incr(vpsci_system_suspend);
+        PSCI_SET_RESULT(regs, do_psci_1_0_system_suspend(epoint, cid));
+        return true;
+    }
+
     default:
         return false;
     }
diff --git a/xen/common/domain.c b/xen/common/domain.c
index 303c338ef2..5b760f0cd9 100644
--- a/xen/common/domain.c
+++ b/xen/common/domain.c
@@ -1326,16 +1326,10 @@ int domain_shutdown(struct domain *d, u8 reason)
     return 0;
 }
 
-void domain_resume(struct domain *d)
+void domain_resume_nopause(struct domain *d)
 {
     struct vcpu *v;
 
-    /*
-     * Some code paths assume that shutdown status does not get reset under
-     * their feet (e.g., some assertions make this assumption).
-     */
-    domain_pause(d);
-
     spin_lock(&d->shutdown_lock);
 
     d->is_shutting_down = d->is_shut_down = 0;
@@ -1349,10 +1343,32 @@ void domain_resume(struct domain *d)
     }
 
     spin_unlock(&d->shutdown_lock);
+}
 
+void domain_resume(struct domain *d)
+{
+    /*
+     * Some code paths assume that shutdown status does not get reset under
+     * their feet (e.g., some assertions make this assumption).
+     */
+    domain_pause(d);
+    domain_resume_nopause(d);
     domain_unpause(d);
 }
 
+bool vcpu_is_suspended(struct vcpu *v)
+{
+    bool is_suspended;
+    struct domain *d = v->domain;
+
+    spin_lock(&d->shutdown_lock);
+    is_suspended = d->shutdown_code == SHUTDOWN_suspend &&
+                   v->paused_for_shutdown;
+    spin_unlock(&d->shutdown_lock);
+
+    return is_suspended;
+}
+
 int vcpu_start_shutdown_deferral(struct vcpu *v)
 {
     if ( v->defer_shutdown )
diff --git a/xen/include/xen/sched.h b/xen/include/xen/sched.h
index fe53d4fab7..2e163497bd 100644
--- a/xen/include/xen/sched.h
+++ b/xen/include/xen/sched.h
@@ -814,6 +814,7 @@ void domain_destroy(struct domain *d);
 int domain_kill(struct domain *d);
 int domain_shutdown(struct domain *d, u8 reason);
 void domain_resume(struct domain *d);
+void domain_resume_nopause(struct domain *d);
 
 int domain_soft_reset(struct domain *d, bool resuming);
 
@@ -824,6 +825,7 @@ int get_domain_state(struct xen_domctl_get_domain_state *info,
 
 int vcpu_start_shutdown_deferral(struct vcpu *v);
 void vcpu_end_shutdown_deferral(struct vcpu *v);
+bool vcpu_is_suspended(struct vcpu *v);
 
 /*
  * Mark specified domain as crashed. This function always returns, even if the
-- 
2.48.1

[PATCH v6 2/4] tools/xl: Allow compilation of 'xl resume' command on Arm

[Mykola Kvach]

From: Mykola Kvach <mykola_kvach@epam.com>

The "xl resume" command was previously excluded from Arm builds because
system suspend/resume (e.g., SYSTEM_SUSPEND via vPSCI) was not
implemented. On x86, this command is used for ACPI S3 suspend/resume.

This change enables compilation of `xl resume` on Arm regardless of the
underlying implementation status, making the tool available for testing
and future feature support. The relevant libxl infrastructure and handler
functions are already present and usable.

The macro `LIBXL_HAVE_NO_SUSPEND_RESUME` has been renamed to
`LIBXL_HAVE_NO_SUSPEND` to better reflect the updated semantics.

Note: This does not imply full system suspend/resume support on Arm.
      The `xl suspend` command still does not work on Arm platforms.

Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
---
Changes in v6:
- Renamed macro from LIBXL_HAVE_NO_SUSPEND_RESUME to LIBXL_HAVE_NO_SUSPEND
  to better reflect the scope of this change
- Applied cosmetic changes based on review feedback
---
 tools/include/libxl.h     |  5 ++---
 tools/xl/xl.h             | 10 +++++-----
 tools/xl/xl_cmdtable.c    |  8 ++++----
 tools/xl/xl_migrate.c     |  4 ++--
 tools/xl/xl_saverestore.c |  4 ++--
 tools/xl/xl_vmcontrol.c   | 14 +++++++-------
 6 files changed, 22 insertions(+), 23 deletions(-)

diff --git a/tools/include/libxl.h b/tools/include/libxl.h
index d6b6e5d2dd..632264912a 100644
--- a/tools/include/libxl.h
+++ b/tools/include/libxl.h
@@ -1128,17 +1128,16 @@ typedef struct libxl__ctx libxl_ctx;
 #define LIBXL_HAVE_SIGCHLD_SHARING 1
 
 /*
- * LIBXL_HAVE_NO_SUSPEND_RESUME
+ * LIBXL_HAVE_NO_SUSPEND
  *
  * Is this is defined then the platform has no support for saving,
  * restoring or migrating a domain. In this case the related functions
  * should be expected to return failure. That is:
  *  - libxl_domain_suspend
- *  - libxl_domain_resume
  *  - libxl_domain_remus_start
  */
 #if defined(__arm__) || defined(__aarch64__)
-#define LIBXL_HAVE_NO_SUSPEND_RESUME 1
+#define LIBXL_HAVE_NO_SUSPEND 1
 #endif
 
 /*
diff --git a/tools/xl/xl.h b/tools/xl/xl.h
index 45745f0dbb..4d4a5bb1c8 100644
--- a/tools/xl/xl.h
+++ b/tools/xl/xl.h
@@ -56,7 +56,7 @@ int create_domain(struct domain_create *dom_info);
 static const char savefileheader_magic[32]=
     "Xen saved domain, xl format\n \0 \r";
 
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
 static const char migrate_receiver_banner[]=
     "xl migration receiver ready, send binary domain data.\n";
 static const char migrate_receiver_ready[]=
@@ -65,7 +65,7 @@ static const char migrate_permission_to_go[]=
     "domain is yours, you are cleared to unpause";
 static const char migrate_report[]=
     "my copy unpause results are as follows";
-#endif
+#endif /* !LIBXL_HAVE_NO_SUSPEND */
 
   /* followed by one byte:
    *     0: everything went well, domain is running
@@ -124,14 +124,14 @@ int main_pciattach(int argc, char **argv);
 int main_pciassignable_add(int argc, char **argv);
 int main_pciassignable_remove(int argc, char **argv);
 int main_pciassignable_list(int argc, char **argv);
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
 int main_restore(int argc, char **argv);
 int main_migrate_receive(int argc, char **argv);
 int main_save(int argc, char **argv);
 int main_migrate(int argc, char **argv);
 int main_suspend(int argc, char **argv);
+#endif /* !LIBXL_HAVE_NO_SUSPEND */
 int main_resume(int argc, char **argv);
-#endif
 int main_dump_core(int argc, char **argv);
 int main_pause(int argc, char **argv);
 int main_unpause(int argc, char **argv);
@@ -202,7 +202,7 @@ int main_cpupoolnumasplit(int argc, char **argv);
 int main_getenforce(int argc, char **argv);
 int main_setenforce(int argc, char **argv);
 int main_loadpolicy(int argc, char **argv);
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
 int main_remus(int argc, char **argv);
 #endif
 int main_devd(int argc, char **argv);
diff --git a/tools/xl/xl_cmdtable.c b/tools/xl/xl_cmdtable.c
index 06a0039718..03f0970bb7 100644
--- a/tools/xl/xl_cmdtable.c
+++ b/tools/xl/xl_cmdtable.c
@@ -152,7 +152,7 @@ const struct cmd_spec cmd_table[] = {
       "                         -autopass\n"
       "--vncviewer-autopass     (consistency alias for --autopass)"
     },
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
     { "save",
       &main_save, 0, 1,
       "Save a domain state to restore later",
@@ -198,12 +198,12 @@ const struct cmd_spec cmd_table[] = {
       "Suspend a domain to RAM",
       "<Domain>",
     },
+#endif /* !LIBXL_HAVE_NO_SUSPEND */
     { "resume",
       &main_resume, 0, 1,
       "Resume a domain from RAM",
       "<Domain>",
     },
-#endif
     { "dump-core",
       &main_dump_core, 0, 1,
       "Core dump a domain",
@@ -524,7 +524,7 @@ const struct cmd_spec cmd_table[] = {
       "Loads a new policy into the Flask Xen security module",
       "<policy file>",
     },
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
     { "remus",
       &main_remus, 0, 1,
       "Enable Remus HA for domain",
@@ -548,7 +548,7 @@ const struct cmd_spec cmd_table[] = {
       "                        checkpoint must be disabled.\n"
       "-p                      Use COLO userspace proxy."
     },
-#endif
+#endif /* !LIBXL_HAVE_NO_SUSPEND */
     { "devd",
       &main_devd, 0, 1,
       "Daemon that listens for devices and launches backends",
diff --git a/tools/xl/xl_migrate.c b/tools/xl/xl_migrate.c
index b8594f44a5..a8e2c39944 100644
--- a/tools/xl/xl_migrate.c
+++ b/tools/xl/xl_migrate.c
@@ -30,7 +30,7 @@
 #include "xl_utils.h"
 #include "xl_parse.h"
 
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
 
 static pid_t create_migration_child(const char *rune, int *send_fd,
                                         int *recv_fd)
@@ -767,7 +767,7 @@ int main_remus(int argc, char **argv)
     close(send_fd);
     return EXIT_FAILURE;
 }
-#endif
+#endif /* !LIBXL_HAVE_NO_SUSPEND */
 
 
 /*
diff --git a/tools/xl/xl_saverestore.c b/tools/xl/xl_saverestore.c
index 953d791d1a..cb10b869b9 100644
--- a/tools/xl/xl_saverestore.c
+++ b/tools/xl/xl_saverestore.c
@@ -29,7 +29,7 @@
 #include "xl_utils.h"
 #include "xl_parse.h"
 
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
 
 void save_domain_core_begin(uint32_t domid,
                             int preserve_domid,
@@ -270,7 +270,7 @@ int main_save(int argc, char **argv)
     return EXIT_SUCCESS;
 }
 
-#endif /* LIBXL_HAVE_NO_SUSPEND_RESUME */
+#endif /* !LIBXL_HAVE_NO_SUSPEND */
 
 
 
diff --git a/tools/xl/xl_vmcontrol.c b/tools/xl/xl_vmcontrol.c
index c813732838..49484ca3e2 100644
--- a/tools/xl/xl_vmcontrol.c
+++ b/tools/xl/xl_vmcontrol.c
@@ -32,17 +32,12 @@
 
 static int fd_lock = -1;
 
-#ifndef LIBXL_HAVE_NO_SUSPEND_RESUME
+#ifndef LIBXL_HAVE_NO_SUSPEND
 static void suspend_domain(uint32_t domid)
 {
     libxl_domain_suspend_only(ctx, domid, NULL);
 }
 
-static void resume_domain(uint32_t domid)
-{
-    libxl_domain_resume(ctx, domid, 1, NULL);
-}
-
 int main_suspend(int argc, char **argv)
 {
     int opt;
@@ -55,6 +50,12 @@ int main_suspend(int argc, char **argv)
 
     return EXIT_SUCCESS;
 }
+#endif /* !LIBXL_HAVE_NO_SUSPEND */
+
+static void resume_domain(uint32_t domid)
+{
+    libxl_domain_resume(ctx, domid, 1, NULL);
+}
 
 int main_resume(int argc, char **argv)
 {
@@ -68,7 +69,6 @@ int main_resume(int argc, char **argv)
 
     return EXIT_SUCCESS;
 }
-#endif
 
 static void pause_domain(uint32_t domid)
 {
-- 
2.48.1

[PATCH v6 3/4] SUPPORT.md: Document PSCI SYSTEM_SUSPEND support for guests

[Mykola Kvach]

From: Mykola Kvach <mykola_kvach@epam.com>

Add a new entry under the "Virtual Hardware, QEMU" section documenting
support for the optional PSCI SYSTEM_SUSPEND function exposed to guests.

This function is available via the virtual PSCI (vPSCI) interface and
allows guest domains (domUs) to initiate system suspend operations.

The feature is currently marked as Experimental.

Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
---
Changes in v6:
- Dropped the generic guest PSCI support entry (merged in a separate patch)
- This patch now documents only the SYSTEM_SUSPEND optional function
- Reworded commit message to match the final form after rebase

Changes in v5:
- Dropped ARM/PSCI entry: this refers to internal use of PSCI SMC calls,
  which is not relevant for SUPPORT.md
- Added a dedicated entry for PSCI SYSTEM_SUSPEND instead of generic guest
  PSCI info; guest PSCI support was documented in a separate patch
---
 SUPPORT.md | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/SUPPORT.md b/SUPPORT.md
index 6a82a92189..b5ab049b52 100644
--- a/SUPPORT.md
+++ b/SUPPORT.md
@@ -962,8 +962,9 @@ Emulated PSCI interface exposed to guests. We support all mandatory
 functions of PSCI 1.1. See below for the list of optional PSCI call
 implemented and their status.
 
-   Status, Mandatory: Supported
-   Status, MIGRATE_INFO_TYPE: Supported
+    Status, Mandatory: Supported
+    Status, MIGRATE_INFO_TYPE: Supported
+    Status, SYSTEM_SUSPEND: Experimental
 
 ## Virtual Hardware, QEMU
 
-- 
2.48.1

[PATCH v6 4/4] CHANGELOG: Document guest suspend/resume to RAM support on Arm

[Mykola Kvach]

From: Mykola Kvach <mykola_kvach@epam.com>

Mention the newly added support for guest suspend and resume to/from
RAM via vPSCI on Arm platforms.

This support is limited to non-hardware domain guests.

Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
Acked-by: Oleksii Kurochko <oleksii.kurochko@gmail.com>
---
Chnages in v6:
- removed reduntand explanation that thi support added for
  both arm32 and arm64.

Changes in v5:
- adjustments to the commit title and message
- expanded the changelog entry to include more context about
  suspend/resume support introduced in this patch series
---
 CHANGELOG.md | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 5f31ca08fe..7e42ca9b59 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -29,6 +29,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/)
 
  - On Arm:
     - Ability to enable stack protector
+    - Support for guest suspend and resume to/from RAM via vPSCI.
+      Applies only to non-hardware domain guests.
 
 ### Removed
  - On x86:
-- 
2.48.1

Re: [PATCH v6 1/4] xen/arm: Implement PSCI SYSTEM_SUSPEND call for guests

[Jan Beulich]

On 24.07.2025 12:40, Andrew Cooper wrote:
> On 24/07/2025 11:08 am, Jan Beulich wrote:
>> On 24.07.2025 11:40, Mykola Kvach wrote:
>>> --- a/xen/common/domain.c
>>> +++ b/xen/common/domain.c
>>> @@ -1326,16 +1326,10 @@ int domain_shutdown(struct domain *d, u8 reason)
>>>      return 0;
>>>  }
>>>  
>>> -void domain_resume(struct domain *d)
>>> +void domain_resume_nopause(struct domain *d)
>> This function not being static on (at least) non-Arm is, I think, in
>> conflict with some Misra rule.
>>
>>>  {
>>>      struct vcpu *v;
>>>  
>>> -    /*
>>> -     * Some code paths assume that shutdown status does not get reset under
>>> -     * their feet (e.g., some assertions make this assumption).
>>> -     */
>>> -    domain_pause(d);
>>> -
>>>      spin_lock(&d->shutdown_lock);
>>>  
>>>      d->is_shutting_down = d->is_shut_down = 0;
>>> @@ -1349,10 +1343,32 @@ void domain_resume(struct domain *d)
>>>      }
>>>  
>>>      spin_unlock(&d->shutdown_lock);
>>> +}
>>>  
>>> +void domain_resume(struct domain *d)
>>> +{
>>> +    /*
>>> +     * Some code paths assume that shutdown status does not get reset under
>>> +     * their feet (e.g., some assertions make this assumption).
>>> +     */
>>> +    domain_pause(d);
>>> +    domain_resume_nopause(d);
>>>      domain_unpause(d);
>>>  }
>>>  
>>> +bool vcpu_is_suspended(struct vcpu *v)
>> pointer-to-const please (and apparently also in do_psci_1_0_system_suspend()).
> 
> No.  This is not a predicate.  It takes a mutable pointer.

Predicate or not doesn't matter. (And it really is a predicate, as the "is" in
the name says, at least to me.)

> (And yes I am irritated that you continue to provide wrong feedback to
> contributors.  I'm in the middle of stripping another bogus const of
> yours from a different area that's hiding a MISRA violation.)

Well, all I know is that we disagree in where we think const is applicable. But
that disagreement was never resolved, and you can't expect me to silently jump
on your train of thinking.

> Also why is this email not on the list?

That I don't know. For sure it wasn't intentional for the list to be dropped.

Jan

Re: [PATCH v6 2/4] tools/xl: Allow compilation of 'xl resume' command on Arm

[Anthony PERARD]

On Thu, Jul 24, 2025 at 12:40:57PM +0300, Mykola Kvach wrote:
> From: Mykola Kvach <mykola_kvach@epam.com>
> 
> The "xl resume" command was previously excluded from Arm builds because
> system suspend/resume (e.g., SYSTEM_SUSPEND via vPSCI) was not
> implemented. On x86, this command is used for ACPI S3 suspend/resume.
> 
> This change enables compilation of `xl resume` on Arm regardless of the
> underlying implementation status, making the tool available for testing
> and future feature support. The relevant libxl infrastructure and handler
> functions are already present and usable.
> 
> The macro `LIBXL_HAVE_NO_SUSPEND_RESUME` has been renamed to
> `LIBXL_HAVE_NO_SUSPEND` to better reflect the updated semantics.
> 
> Note: This does not imply full system suspend/resume support on Arm.
>       The `xl suspend` command still does not work on Arm platforms.
> 
> Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
> ---
> Changes in v6:
> - Renamed macro from LIBXL_HAVE_NO_SUSPEND_RESUME to LIBXL_HAVE_NO_SUSPEND
>   to better reflect the scope of this change
> - Applied cosmetic changes based on review feedback
> ---
>  tools/include/libxl.h     |  5 ++---
>  tools/xl/xl.h             | 10 +++++-----
>  tools/xl/xl_cmdtable.c    |  8 ++++----
>  tools/xl/xl_migrate.c     |  4 ++--
>  tools/xl/xl_saverestore.c |  4 ++--
>  tools/xl/xl_vmcontrol.c   | 14 +++++++-------
>  6 files changed, 22 insertions(+), 23 deletions(-)
> 
> diff --git a/tools/include/libxl.h b/tools/include/libxl.h
> index d6b6e5d2dd..632264912a 100644
> --- a/tools/include/libxl.h
> +++ b/tools/include/libxl.h
> @@ -1128,17 +1128,16 @@ typedef struct libxl__ctx libxl_ctx;
>  #define LIBXL_HAVE_SIGCHLD_SHARING 1
>  
>  /*
> - * LIBXL_HAVE_NO_SUSPEND_RESUME
> + * LIBXL_HAVE_NO_SUSPEND
>   *
>   * Is this is defined then the platform has no support for saving,
>   * restoring or migrating a domain. In this case the related functions
>   * should be expected to return failure. That is:
>   *  - libxl_domain_suspend
> - *  - libxl_domain_resume
>   *  - libxl_domain_remus_start
>   */
>  #if defined(__arm__) || defined(__aarch64__)
> -#define LIBXL_HAVE_NO_SUSPEND_RESUME 1
> +#define LIBXL_HAVE_NO_SUSPEND 1
>  #endif

I'm sorry, if you remove LIBXL_HAVE_NO_SUSPEND_RESUME, you have to
implement all the function listed. I'm pretty sure `libvirt` isn't going
to build (on arm) if you remove that macro... Actually, libvirt is going
to build, it's going to expect migration to work, and probably allow to
try to migrate Arm VMs instead of bailing out early.

I wonder what this LIBXL_HAVE_NO_SUSPEND_RESUME is for, since you don't
make any changes to libxl (tools/libs/light), but only to program that
make use of it.

Looking at 3ac3817762d1 ("xl: suppress suspend/resume functions on platforms which do not support it.")
    https://xenbits.xenproject.org/gitweb/?p=xen.git;a=commitdiff;h=3ac3817762d1a8b39fa45998ec8c40cabfcfc802
it seems the real purpose was just an hint that migrate/suspend/saving
aren't going to work on that platform.

Looks like `xl resume` is a fairly new command which only makes use if
libxl_domain_resume outside of migration, but the macro
LIBXL_HAVE_NO_SUSPEND_RESUME was mostly a hint that migration doesn't
work.

So I think moving the `xl resume` command out of
LIBXL_HAVE_NO_SUSPEND_RESUME would be good enough for this patch,
without touching libxl.h.

Cheers,

-- 
Anthony PERARD

Re: [PATCH v6 2/4] tools/xl: Allow compilation of 'xl resume' command on Arm

[Mykola Kvach]

Hi Anthony,

On Thu, Jul 24, 2025 at 5:01 PM Anthony PERARD <anthony@xenproject.org> wrote:
>
> On Thu, Jul 24, 2025 at 12:40:57PM +0300, Mykola Kvach wrote:
> > From: Mykola Kvach <mykola_kvach@epam.com>
> >
> > The "xl resume" command was previously excluded from Arm builds because
> > system suspend/resume (e.g., SYSTEM_SUSPEND via vPSCI) was not
> > implemented. On x86, this command is used for ACPI S3 suspend/resume.
> >
> > This change enables compilation of `xl resume` on Arm regardless of the
> > underlying implementation status, making the tool available for testing
> > and future feature support. The relevant libxl infrastructure and handler
> > functions are already present and usable.
> >
> > The macro `LIBXL_HAVE_NO_SUSPEND_RESUME` has been renamed to
> > `LIBXL_HAVE_NO_SUSPEND` to better reflect the updated semantics.
> >
> > Note: This does not imply full system suspend/resume support on Arm.
> >       The `xl suspend` command still does not work on Arm platforms.
> >
> > Signed-off-by: Mykola Kvach <mykola_kvach@epam.com>
> > ---
> > Changes in v6:
> > - Renamed macro from LIBXL_HAVE_NO_SUSPEND_RESUME to LIBXL_HAVE_NO_SUSPEND
> >   to better reflect the scope of this change
> > - Applied cosmetic changes based on review feedback
> > ---
> >  tools/include/libxl.h     |  5 ++---
> >  tools/xl/xl.h             | 10 +++++-----
> >  tools/xl/xl_cmdtable.c    |  8 ++++----
> >  tools/xl/xl_migrate.c     |  4 ++--
> >  tools/xl/xl_saverestore.c |  4 ++--
> >  tools/xl/xl_vmcontrol.c   | 14 +++++++-------
> >  6 files changed, 22 insertions(+), 23 deletions(-)
> >
> > diff --git a/tools/include/libxl.h b/tools/include/libxl.h
> > index d6b6e5d2dd..632264912a 100644
> > --- a/tools/include/libxl.h
> > +++ b/tools/include/libxl.h
> > @@ -1128,17 +1128,16 @@ typedef struct libxl__ctx libxl_ctx;
> >  #define LIBXL_HAVE_SIGCHLD_SHARING 1
> >
> >  /*
> > - * LIBXL_HAVE_NO_SUSPEND_RESUME
> > + * LIBXL_HAVE_NO_SUSPEND
> >   *
> >   * Is this is defined then the platform has no support for saving,
> >   * restoring or migrating a domain. In this case the related functions
> >   * should be expected to return failure. That is:
> >   *  - libxl_domain_suspend
> > - *  - libxl_domain_resume
> >   *  - libxl_domain_remus_start
> >   */
> >  #if defined(__arm__) || defined(__aarch64__)
> > -#define LIBXL_HAVE_NO_SUSPEND_RESUME 1
> > +#define LIBXL_HAVE_NO_SUSPEND 1
> >  #endif
>
> I'm sorry, if you remove LIBXL_HAVE_NO_SUSPEND_RESUME, you have to
> implement all the function listed. I'm pretty sure `libvirt` isn't going
> to build (on arm) if you remove that macro... Actually, libvirt is going
> to build, it's going to expect migration to work, and probably allow to
> try to migrate Arm VMs instead of bailing out early.
>
> I wonder what this LIBXL_HAVE_NO_SUSPEND_RESUME is for, since you don't
> make any changes to libxl (tools/libs/light), but only to program that
> make use of it.
>
> Looking at 3ac3817762d1 ("xl: suppress suspend/resume functions on platforms which do not support it.")
>     https://xenbits.xenproject.org/gitweb/?p=xen.git;a=commitdiff;h=3ac3817762d1a8b39fa45998ec8c40cabfcfc802
> it seems the real purpose was just an hint that migrate/suspend/saving
> aren't going to work on that platform.
>
> Looks like `xl resume` is a fairly new command which only makes use if
> libxl_domain_resume outside of migration, but the macro
> LIBXL_HAVE_NO_SUSPEND_RESUME was mostly a hint that migration doesn't
> work.
>
> So I think moving the `xl resume` command out of
> LIBXL_HAVE_NO_SUSPEND_RESUME would be good enough for this patch,
> without touching libxl.h.

Got it. I'll revert the patch to the previous version.
Thanks for the review and clarification -- much appreciated!

>
> Cheers,
>
> --
> Anthony PERARD

Best regards,
Mykola