[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[PATCH v5 1/4] xen/arm: Implement PSCI SYSTEM_SUSPEND call for guests



From: Mykola Kvach <mykola_kvach@xxxxxxxx>

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
- the domain is shut down with SHUTDOWN_suspend, and resumes execution at
  the address provided during suspend by guest

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@xxxxxxxx>
---
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/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/mmu/p2m.c                |  6 +-
 xen/arch/arm/vpsci.c                  | 96 +++++++++++++++++++++++----
 5 files changed, 92 insertions(+), 15 deletions(-)

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/mmu/p2m.c b/xen/arch/arm/mmu/p2m.c
index 67296dabb5..f9c09a49e2 100644
--- a/xen/arch/arm/mmu/p2m.c
+++ b/xen/arch/arm/mmu/p2m.c
@@ -6,6 +6,8 @@
 #include <xen/sched.h>
 #include <xen/softirq.h>
 
+#include <public/sched.h>
+
 #include <asm/alternative.h>
 #include <asm/event.h>
 #include <asm/flushtlb.h>
@@ -198,7 +200,9 @@ void dump_p2m_lookup(struct domain *d, paddr_t addr)
  */
 void p2m_save_state(struct vcpu *p)
 {
-    p->arch.sctlr = READ_SYSREG(SCTLR_EL1);
+    if ( !(p->domain->is_shutting_down &&
+           p->domain->shutdown_code == SHUTDOWN_suspend) )
+        p->arch.sctlr = READ_SYSREG(SCTLR_EL1);
 
     if ( cpus_have_const_cap(ARM64_WORKAROUND_AT_SPECULATE) )
     {
diff --git a/xen/arch/arm/vpsci.c b/xen/arch/arm/vpsci.c
index 7ba9ccd94b..c6b9ac1fc8 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 +208,52 @@ static void do_psci_0_2_system_reset(void)
     domain_shutdown(d,SHUTDOWN_reboot);
 }
 
+static void do_resume_on_error(struct domain *d)
+{
+    struct vcpu *v;
+
+    spin_lock(&d->shutdown_lock);
+
+    d->is_shutting_down = d->is_shut_down = 0;
+    d->shutdown_code = SHUTDOWN_CODE_INVALID;
+
+    for_each_vcpu ( d, v )
+    {
+        if ( v->paused_for_shutdown )
+            vcpu_unpause(v);
+        v->paused_for_shutdown = 0;
+    }
+
+    spin_unlock(&d->shutdown_lock);
+}
+
+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;
+
+    /* Drop this check once SYSTEM_SUSPEND is supported in hardware domain */
+    if ( is_hardware_domain(d) )
+        return PSCI_NOT_SUPPORTED;
+
+    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) )
+        {
+            do_resume_on_error(d);
+            return PSCI_DENIED;
+        }
+
+    ret = do_setup_vcpu_ctx(current, epoint, cid);
+    if ( ret != PSCI_SUCCESS )
+        do_resume_on_error(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 +271,8 @@ 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 +403,17 @@ 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);
+
+        perfc_incr(vpsci_system_suspend);
+        PSCI_SET_RESULT(regs, do_psci_1_0_system_suspend(epoint, cid));
+        return true;
+    }
+
     default:
         return false;
     }
-- 
2.48.1




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.