Skip to contentSkip to conversation

ppc/spapr: Support reboot of secure pseries guest

PATCH Superseded Bharata B Rao 3 participants 12 messages Dec 9, 2019 – Dec 12, 2019

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

cover ppc/spapr: Support reboot of secure pseries guest

This patchset adds KVM_PPC_SVM_OFF ioctl which is required to support reset of secure guest. This includes linux-headers update so that we get the newly introduced ioctl.

v0: https://lists.gnu.org/archive/html/qemu-devel/2019-07/msg02408.html

Bharata B Rao (2):
linux-headers: Update
ppc/spapr: Support reboot of secure pseries guest

 hw/ppc/spapr.c                                |  1 +
 include/standard-headers/asm-x86/bootparam.h  |  7 +-
 .../infiniband/hw/vmw_pvrdma/pvrdma_dev_api.h | 15 +++-
 include/standard-headers/drm/drm_fourcc.h     | 28 ++++++-
 .../linux/input-event-codes.h                 | 77 +++++++++++++++++++
 include/standard-headers/linux/pci_regs.h     |  3 +
 .../standard-headers/rdma/vmw_pvrdma-abi.h    |  5 ++
 linux-headers/linux/kvm.h                     |  1 +
 target/ppc/kvm.c                              |  7 ++
 target/ppc/kvm_ppc.h                          |  6 ++
 10 files changed, 146 insertions(+), 4 deletions(-)
Signature · 1 line
2.21.0
#1 lore raw

PATCH 1/2 linux-headers: Update

Update to mainline commit: e42617b825f8 ("Linux 5.5-rc1")

---
 include/standard-headers/asm-x86/bootparam.h  |  7 +-
 .../infiniband/hw/vmw_pvrdma/pvrdma_dev_api.h | 15 +++-
 include/standard-headers/drm/drm_fourcc.h     | 28 ++++++-
 .../linux/input-event-codes.h                 | 77 +++++++++++++++++++
 include/standard-headers/linux/pci_regs.h     |  3 +
 .../standard-headers/rdma/vmw_pvrdma-abi.h    |  5 ++
 linux-headers/linux/kvm.h                     |  1 +
 7 files changed, 132 insertions(+), 4 deletions(-)
include/standard-headers/asm-x86/bootparam.h +6 -1
index a6f7cf535e..072e2ed546 100644
@@ -2,7 +2,7 @@
 #ifndef _ASM_X86_BOOTPARAM_H
 #define _ASM_X86_BOOTPARAM_H
 
-/* setup_data types */
+/* setup_data/setup_indirect types */
 #define SETUP_NONE			0
 #define SETUP_E820_EXT			1
 #define SETUP_DTB			2
@@ -11,6 +11,11 @@
 #define SETUP_APPLE_PROPERTIES		5
 #define SETUP_JAILHOUSE			6
 
+#define SETUP_INDIRECT			(1<<31)
+
+/* SETUP_INDIRECT | max(SETUP_*) */
+#define SETUP_TYPE_MAX			(SETUP_INDIRECT | SETUP_JAILHOUSE)
+
 /* ram_size flags */
 #define RAMDISK_IMAGE_START_MASK	0x07FF
 #define RAMDISK_PROMPT_FLAG		0x8000
include/standard-headers/drivers/infiniband/hw/vmw_pvrdma/pvrdma_dev_api.h +14 -1
index d019872608..a5a1c8234e 100644
@@ -58,7 +58,8 @@
 #define PVRDMA_ROCEV1_VERSION		17
 #define PVRDMA_ROCEV2_VERSION		18
 #define PVRDMA_PPN64_VERSION		19
-#define PVRDMA_VERSION			PVRDMA_PPN64_VERSION
+#define PVRDMA_QPHANDLE_VERSION		20
+#define PVRDMA_VERSION			PVRDMA_QPHANDLE_VERSION
 
 #define PVRDMA_BOARD_ID			1
 #define PVRDMA_REV_ID			1
@@ -581,6 +582,17 @@ struct pvrdma_cmd_create_qp_resp {
 	uint32_t max_inline_data;
 };
 
+struct pvrdma_cmd_create_qp_resp_v2 {
+	struct pvrdma_cmd_resp_hdr hdr;
+	uint32_t qpn;
+	uint32_t qp_handle;
+	uint32_t max_send_wr;
+	uint32_t max_recv_wr;
+	uint32_t max_send_sge;
+	uint32_t max_recv_sge;
+	uint32_t max_inline_data;
+};
+
 struct pvrdma_cmd_modify_qp {
 	struct pvrdma_cmd_hdr hdr;
 	uint32_t qp_handle;
@@ -663,6 +675,7 @@ union pvrdma_cmd_resp {
 	struct pvrdma_cmd_create_cq_resp create_cq_resp;
 	struct pvrdma_cmd_resize_cq_resp resize_cq_resp;
 	struct pvrdma_cmd_create_qp_resp create_qp_resp;
+	struct pvrdma_cmd_create_qp_resp_v2 create_qp_resp_v2;
 	struct pvrdma_cmd_query_qp_resp query_qp_resp;
 	struct pvrdma_cmd_destroy_qp_resp destroy_qp_resp;
 	struct pvrdma_cmd_create_srq_resp create_srq_resp;
include/standard-headers/drm/drm_fourcc.h +26 -2
index a308c91b4f..46d279f515 100644
@@ -68,7 +68,7 @@ extern "C" {
 #define fourcc_code(a, b, c, d) ((uint32_t)(a) | ((uint32_t)(b) << 8) | \
 				 ((uint32_t)(c) << 16) | ((uint32_t)(d) << 24))
 
-#define DRM_FORMAT_BIG_ENDIAN (1<<31) /* format is big endian instead of little endian */
+#define DRM_FORMAT_BIG_ENDIAN (1U<<31) /* format is big endian instead of little endian */
 
 /* Reserve 0 for the invalid format specifier */
 #define DRM_FORMAT_INVALID	0
@@ -647,7 +647,21 @@ extern "C" {
  * Further information on the use of AFBC modifiers can be found in
  * Documentation/gpu/afbc.rst
  */
-#define DRM_FORMAT_MOD_ARM_AFBC(__afbc_mode)	fourcc_mod_code(ARM, __afbc_mode)
+
+/*
+ * The top 4 bits (out of the 56 bits alloted for specifying vendor specific
+ * modifiers) denote the category for modifiers. Currently we have only two
+ * categories of modifiers ie AFBC and MISC. We can have a maximum of sixteen
+ * different categories.
+ */
+#define DRM_FORMAT_MOD_ARM_CODE(__type, __val) \
+	fourcc_mod_code(ARM, ((uint64_t)(__type) << 52) | ((__val) & 0x000fffffffffffffULL))
+
+#define DRM_FORMAT_MOD_ARM_TYPE_AFBC 0x00
+#define DRM_FORMAT_MOD_ARM_TYPE_MISC 0x01
+
+#define DRM_FORMAT_MOD_ARM_AFBC(__afbc_mode) \
+	DRM_FORMAT_MOD_ARM_CODE(DRM_FORMAT_MOD_ARM_TYPE_AFBC, __afbc_mode)
 
 /*
  * AFBC superblock size
@@ -741,6 +755,16 @@ extern "C" {
  */
 #define AFBC_FORMAT_MOD_BCH     (1ULL << 11)
 
+/*
+ * Arm 16x16 Block U-Interleaved modifier
+ *
+ * This is used by Arm Mali Utgard and Midgard GPUs. It divides the image
+ * into 16x16 pixel blocks. Blocks are stored linearly in order, but pixels
+ * in the block are reordered.
+ */
+#define DRM_FORMAT_MOD_ARM_16X16_BLOCK_U_INTERLEAVED \
+	DRM_FORMAT_MOD_ARM_CODE(DRM_FORMAT_MOD_ARM_TYPE_MISC, 1ULL)
+
 /*
  * Allwinner tiled modifier
  *
include/standard-headers/linux/input-event-codes.h +77 -0

83 lines not on this page · raw message

include/standard-headers/linux/pci_regs.h +3 -0
index 29d6e93fd1..acb7d2bdb4 100644
@@ -34,6 +34,7 @@
  * of which the first 64 bytes are standardized as follows:
  */
 #define PCI_STD_HEADER_SIZEOF	64
+#define PCI_STD_NUM_BARS	6	/* Number of standard BARs */
 #define PCI_VENDOR_ID		0x00	/* 16 bits */
 #define PCI_DEVICE_ID		0x02	/* 16 bits */
 #define PCI_COMMAND		0x04	/* 16 bits */
@@ -673,6 +674,8 @@
 #define  PCI_EXP_LNKCTL2_TLS_8_0GT	0x0003 /* Supported Speed 8GT/s */
 #define  PCI_EXP_LNKCTL2_TLS_16_0GT	0x0004 /* Supported Speed 16GT/s */
 #define  PCI_EXP_LNKCTL2_TLS_32_0GT	0x0005 /* Supported Speed 32GT/s */
+#define  PCI_EXP_LNKCTL2_ENTER_COMP	0x0010 /* Enter Compliance */
+#define  PCI_EXP_LNKCTL2_TX_MARGIN	0x0380 /* Transmit Margin */
 #define PCI_EXP_LNKSTA2		50	/* Link Status 2 */
 #define PCI_CAP_EXP_ENDPOINT_SIZEOF_V2	52	/* v2 endpoints with link end here */
 #define PCI_EXP_SLTCAP2		52	/* Slot Capabilities 2 */
include/standard-headers/rdma/vmw_pvrdma-abi.h +5 -0
index 336a8d596f..0989426a3f 100644
@@ -179,6 +179,11 @@ struct pvrdma_create_qp {
 	uint64_t __attribute__((aligned(8))) qp_addr;
 };
 
+struct pvrdma_create_qp_resp {
+	uint32_t qpn;
+	uint32_t qp_handle;
+};
+
 /* PVRDMA masked atomic compare and swap */
 struct pvrdma_ex_cmp_swap {
 	uint64_t __attribute__((aligned(8))) swap_val;
linux-headers/linux/kvm.h +1 -0
index 3b27a1ae85..9d647fad76 100644
@@ -1348,6 +1348,7 @@ struct kvm_s390_ucas_mapping {
 #define KVM_PPC_GET_CPU_CHAR	  _IOR(KVMIO,  0xb1, struct kvm_ppc_cpu_char)
 /* Available with KVM_CAP_PMU_EVENT_FILTER */
 #define KVM_SET_PMU_EVENT_FILTER  _IOW(KVMIO,  0xb2, struct kvm_pmu_event_filter)
+#define KVM_PPC_SVM_OFF		  _IO(KVMIO,  0xb3)
 
 /* ioctl for vm fd */
 #define KVM_CREATE_DEVICE	  _IOWR(KVMIO,  0xe0, struct kvm_create_device)
Signature · 1 line
2.21.0
#2 lore raw

PATCH 2/2 ppc/spapr: Support reboot of secure pseries guest

In reply to Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +1 -0
index f11422fc41..4c7ad3400d 100644
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
+    kvmppc_svm_off();
     spapr_caps_apply(spapr);
 
     first_ppc_cpu = POWERPC_CPU(first_cpu);
target/ppc/kvm.c +7 -0
index 7406d18945..1a86fa4f0c 100644
@@ -2900,3 +2900,10 @@ void kvmppc_set_reg_tb_offset(PowerPCCPU *cpu, int64_t tb_offset)
         kvm_set_one_reg(cs, KVM_REG_PPC_TB_OFFSET, &tb_offset);
     }
 }
+
+int kvmppc_svm_off(void)
+{
+    KVMState *s = KVM_STATE(current_machine->accelerator);
+
+    return kvm_vm_ioctl(s, KVM_PPC_SVM_OFF);
+}
target/ppc/kvm_ppc.h +6 -0
index 47b08a4030..5cc812e486 100644
@@ -37,6 +37,7 @@ int kvmppc_booke_watchdog_enable(PowerPCCPU *cpu);
 target_ulong kvmppc_configure_v3_mmu(PowerPCCPU *cpu,
                                      bool radix, bool gtse,
                                      uint64_t proc_tbl);
+int kvmppc_svm_off(void);
 #ifndef CONFIG_USER_ONLY
 bool kvmppc_spapr_use_multitce(void);
 int kvmppc_spapr_enable_inkernel_multitce(void);
@@ -201,6 +202,11 @@ static inline target_ulong kvmppc_configure_v3_mmu(PowerPCCPU *cpu,
     return 0;
 }
 
+static inline int kvmppc_svm_off(void)
+{
+    return 0;
+}
+
 static inline void kvmppc_set_reg_ppc_online(PowerPCCPU *cpu,
                                              unsigned int online)
 {
Signature · 1 line
2.21.0
#3 lore raw
Quoted 24 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 4 lines
hw/ppc/spapr.c +1 -0
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

Quoted 42 lines
hw/ppc/spapr.c +0 -0
     spapr_caps_apply(spapr);
 
     first_ppc_cpu = POWERPC_CPU(first_cpu);
target/ppc/kvm.c +7 -0
index 7406d18945..1a86fa4f0c 100644
@@ -2900,3 +2900,10 @@ void kvmppc_set_reg_tb_offset(PowerPCCPU *cpu, int64_t tb_offset)
         kvm_set_one_reg(cs, KVM_REG_PPC_TB_OFFSET, &tb_offset);
     }
 }
+
+int kvmppc_svm_off(void)
+{
+    KVMState *s = KVM_STATE(current_machine->accelerator);
+
+    return kvm_vm_ioctl(s, KVM_PPC_SVM_OFF);
+}
target/ppc/kvm_ppc.h +6 -0
index 47b08a4030..5cc812e486 100644
@@ -37,6 +37,7 @@ int kvmppc_booke_watchdog_enable(PowerPCCPU *cpu);
 target_ulong kvmppc_configure_v3_mmu(PowerPCCPU *cpu,
                                      bool radix, bool gtse,
                                      uint64_t proc_tbl);
+int kvmppc_svm_off(void);
 #ifndef CONFIG_USER_ONLY
 bool kvmppc_spapr_use_multitce(void);
 int kvmppc_spapr_enable_inkernel_multitce(void);
@@ -201,6 +202,11 @@ static inline target_ulong kvmppc_configure_v3_mmu(PowerPCCPU *cpu,
     return 0;
 }
 
+static inline int kvmppc_svm_off(void)
+{
+    return 0;
+}
+
 static inline void kvmppc_set_reg_ppc_online(PowerPCCPU *cpu,
                                              unsigned int online)
 {
Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson
#4 lore raw
Quoted 28 lines · David Gibson
Quoted 27 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
Quoted 4 lines

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 1 line
hw/ppc/spapr.c +1 -0
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

I could, by spapr_machine_reset() and the callers don't propagate the errors up. So may be I could print a warning instead when ioctl fails?

Regards,
Bharata.

#5 lore raw
Quoted 29 lines · Bharata B Rao
Quoted 28 lines · David Gibson
Quoted 27 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
Quoted 7 lines
Quoted 4 lines

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 1 line
hw/ppc/spapr.c +1 -0
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

I could, by spapr_machine_reset() and the callers don't propagate the errors up. So may be I could print a warning instead when ioctl fails?

An error here means you cannot restart the machine and should probably suspend, or try until it is not EBUSY (==all threads have stopped?).

Signature · 1 line
Alexey
#6 lore raw
Quoted 37 lines · Alexey Kardashevskiy
Quoted 34 lines · Bharata B Rao
Quoted 32 lines · David Gibson
Quoted 24 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 4 lines
hw/ppc/spapr.c +1 -0
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

Quoted 5 lines
Quoted 2 lines

I could, by spapr_machine_reset() and the callers don't propagate the errors up. So may be I could print a warning instead when ioctl fails?

An error here means you cannot restart the machine and should probably suspend, or try until it is not EBUSY (==all threads have stopped?).

Right, if this fails, something has gone badly wrong. You should absolutely print a message, and in fact it might be appropriate to quit outright. IIUC the way PEF resets work, a failure here means you won't be able to boot after the reset, since the guest memory will still be inaccessible to the host.

Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson
#7 lore raw
Quoted 41 lines · David Gibson
Quoted 40 lines · Alexey Kardashevskiy
Quoted 36 lines · Bharata B Rao
Quoted 28 lines · David Gibson
Quoted 27 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
Quoted 4 lines

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 1 line
hw/ppc/spapr.c +1 -0
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

I could, by spapr_machine_reset() and the callers don't propagate the errors up. So may be I could print a warning instead when ioctl fails?

Quoted 8 lines
Quoted 2 lines

An error here means you cannot restart the machine and should probably suspend, or try until it is not EBUSY (==all threads have stopped?).

Right, if this fails, something has gone badly wrong. You should absolutely print a message, and in fact it might be appropriate to quit outright. IIUC the way PEF resets work, a failure here means you won't be able to boot after the reset, since the guest memory will still be inaccessible to the host.

Correct. I will send next version with a message and abort() added in the ioctl failure path.

Regards,
Bharata.

#8 lore raw
Quoted 45 lines · Bharata B Rao
Quoted 44 lines · David Gibson
Quoted 42 lines · Alexey Kardashevskiy
Quoted 29 lines · Bharata B Rao
Quoted 28 lines · David Gibson
Quoted 27 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
Quoted 7 lines
Quoted 4 lines

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 1 line
hw/ppc/spapr.c +1 -0
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

I could, by spapr_machine_reset() and the callers don't propagate the errors up. So may be I could print a warning instead when ioctl fails?

An error here means you cannot restart the machine and should probably suspend, or try until it is not EBUSY (==all threads have stopped?).

Quoted 8 lines
Quoted 5 lines

Right, if this fails, something has gone badly wrong. You should absolutely print a message, and in fact it might be appropriate to quit outright. IIUC the way PEF resets work, a failure here means you won't be able to boot after the reset, since the guest memory will still be inaccessible to the host.

Correct. I will send next version with a message and abort() added in the ioctl failure path.

abort() or assert() isn't right either - that's reserved for things that are definitely caused by a qemu code bug. This should be an exit(EXIT_FAILURE).

Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson
#9 lore raw
Quoted 52 lines · David Gibson
Quoted 51 lines · Bharata B Rao
Quoted 49 lines · David Gibson
Quoted 37 lines · Alexey Kardashevskiy
Quoted 34 lines · Bharata B Rao
Quoted 32 lines · David Gibson
Quoted 24 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 4 lines
hw/ppc/spapr.c +1 -0
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

Quoted 5 lines
Quoted 2 lines

I could, by spapr_machine_reset() and the callers don't propagate the errors up. So may be I could print a warning instead when ioctl fails?

An error here means you cannot restart the machine and should probably suspend, or try until it is not EBUSY (==all threads have stopped?).

Right, if this fails, something has gone badly wrong. You should absolutely print a message, and in fact it might be appropriate to quit outright. IIUC the way PEF resets work, a failure here means you won't be able to boot after the reset, since the guest memory will still be inaccessible to the host.

Quoted 6 lines
Quoted 2 lines

Correct. I will send next version with a message and abort() added in the ioctl failure path.

abort() or assert() isn't right either - that's reserved for things that are definitely caused by a qemu code bug. This should be an exit(EXIT_FAILURE).

Ok, but I see a problem with checking the return value of this ioctl from userspace. If this ioctl is run on older kernels that don't support this ioctl, we get -ENOTTY as return value. We shouldn't be exiting in that case.

It looks like we may need a new KVM capability to advertise the presence of KVM_PPC_SVM_OFF ioctl (or more generally, to advertise host kernel's capability to support secure guests).

Paul - Do you think we should add such a KVM capability? Here is the summary of the problem:

1. QEMU invokes KVM_PPC_SVM_OFF ioctl from machine reset path and currently we don't check for its return value.
2. On host kernels that support secure guests,
2a. this ioctl returns 0 for regular guests and has no effect. 2b. the ioctl can fail for secure guests and here we could check the
return value and exit the guest right away.
3. On host kernels that don't support secure guests, ioctl returns -ENOTTY but we ignore the return value and continue the reset as this is for a non-secure guest.

If we have such a KVM capability, we could invoke the ioctl only if it is supported and handle the return value appropriately.

Regards,
Bharata.

#10 lore raw
Quoted 56 lines · Bharata B Rao
Quoted 55 lines · David Gibson
Quoted 53 lines · Bharata B Rao
Quoted 41 lines · David Gibson
Quoted 40 lines · Alexey Kardashevskiy
Quoted 36 lines · Bharata B Rao
Quoted 28 lines · David Gibson
Quoted 27 lines · Bharata B Rao

A pseries guest can be run as a secure guest on Ultravisor-enabled POWER platforms. When such a secure guest is reset, we need to release/reset a few resources both on ultravisor and hypervisor side. This is achieved by invoking this new ioctl KVM_PPC_SVM_OFF from the machine reset path.

As part of this ioctl, the secure guest is essentially transitioned back to normal mode so that it can reboot like a regular guest and become secure again.

This ioctl has no effect when invoked for a normal guest.

---
 hw/ppc/spapr.c       | 1 +
 target/ppc/kvm.c     | 7 +++++++
 target/ppc/kvm_ppc.h | 6 ++++++
 3 files changed, 14 insertions(+)
hw/ppc/spapr.c +0 -0
index f11422fc41..4c7ad3400d 100644
@@ -1597,6 +1597,7 @@ static void spapr_machine_reset(MachineState *machine)
     void *fdt;
     int rc;
 
Quoted 4 lines

hw/ppc/spapr.c static void spapr_machine_reset(MachineState *machine) · line 1600

Quoted 1 line
hw/ppc/spapr.c +1 -0
+    kvmppc_svm_off();

If you're going to have this return an error value, you should really check it here.

I could, by spapr_machine_reset() and the callers don't propagate the errors up. So may be I could print a warning instead when ioctl fails?

Quoted 8 lines
Quoted 2 lines

An error here means you cannot restart the machine and should probably suspend, or try until it is not EBUSY (==all threads have stopped?).

Right, if this fails, something has gone badly wrong. You should absolutely print a message, and in fact it might be appropriate to quit outright. IIUC the way PEF resets work, a failure here means you won't be able to boot after the reset, since the guest memory will still be inaccessible to the host.

Correct. I will send next version with a message and abort() added in the ioctl failure path.

Quoted 8 lines
Quoted 3 lines

abort() or assert() isn't right either - that's reserved for things that are definitely caused by a qemu code bug. This should be an exit(EXIT_FAILURE).

Ok, but I see a problem with checking the return value of this ioctl from userspace. If this ioctl is run on older kernels that don't support this ioctl, we get -ENOTTY as return value. We shouldn't be exiting in that case.

Ah, right. We'll need to check for -ENOTTY specifically and ignore it, then. We don't want this spewing warnings on every non-secure guest.

Quoted 3 lines

It looks like we may need a new KVM capability to advertise the presence of KVM_PPC_SVM_OFF ioctl (or more generally, to advertise host kernel's capability to support secure guests).

Actually, that's probably a better idea still.

Quoted 19 lines

Paul - Do you think we should add such a KVM capability? Here is the summary of the problem:

1. QEMU invokes KVM_PPC_SVM_OFF ioctl from machine reset path and currently we don't check for its return value.
2. On host kernels that support secure guests,
2a. this ioctl returns 0 for regular guests and has no effect. 2b. the ioctl can fail for secure guests and here we could check the
return value and exit the guest right away.
3. On host kernels that don't support secure guests, ioctl returns -ENOTTY but we ignore the return value and continue the reset as this is for a non-secure guest.

If we have such a KVM capability, we could invoke the ioctl only if it is supported and handle the return value appropriately.

Regards,
Bharata.

Signature · 4 lines
David Gibson			| I'll have my music baroque, and my code
david AT gibson.dropbear.id.au	| minimalist, thank you.  NOT _the_ _other_
				| _way_ _around_!
http://www.ozlabs.org/~dgibson
#11 lore raw
Quoted 3 lines · David Gibson

Ah, right. We'll need to check for -ENOTTY specifically and ignore it, then. We don't want this spewing warnings on every non-secure guest.

I am posting v2 with explicit check for -ENOTTY.

Quoted 6 lines
Quoted 4 lines

It looks like we may need a new KVM capability to advertise the presence of KVM_PPC_SVM_OFF ioctl (or more generally, to advertise host kernel's capability to support secure guests).

Actually, that's probably a better idea still.

If and when we decide to have this KVM capability and that goes upstream, we can update the QEMU accordingly?

Regards,
Bharata.