Skip to contentSkip to conversation

ISA 3.00 KVM guest support

RFC PATCH In review Sam Bobroff 4 participants 28 messages Feb 23, 2017 – Mar 3, 2017

lore
Share & tools
lore
#0 lore raw

v2 cover ISA 3.00 KVM guest support

Update notes:

Since the last version, there has been a change to specification of the values used during client architecture support, regarding the bits in option vector 5, so some bits and other processing have changed a bit.

This version has not been as well tested as the last. Testing is ongoing.

General intro:

Because KVM will soon provide the necessary infrastructure for KVM guests to run on POWER9 CPUs, we can now start exploiting this new functionality from QEMU. See:
https://lists.ozlabs.org/pipermail/linuxppc-dev/2017-January/153433.html

This work is not yet complete but it is functional and is presented for early review. It overlaps in some places with current work supporting the same guests under full emulation.

This set aims to support only the following scenarios:
* A POWER9 host running in radix mode, running a guest in radix mode.
* A POWER9 host running in hash mode, running a guest in hash mode.
* A POWER9 host running in hash mode, running a guest in legacy(+) mode. (+) Where legacy means that the guest does not support ISA 3.00.

Hash or radix mode for the host is controlled via the "disable_radix" kernel command line parameter: the host will use radix unless disable_radix is given. For the guest it should be automatically selected to match the host.

Bad legacy guests: There are some recent kernels (e.g. 4.9) that will, when run as a KVM guest and if the ibm,pa-features entry in the device tree has the Radix MMU bit set, attempt to initialize the MMU as if they were a host (which will cause them to crash). To avoid exposing this problem, the Radix MMU bit is removed from ibm,pa-features when a legacy guest is detected.

Final Notes:
* Migration/snapshots are not yet investigated.
* This set is based on the ppc-for-2.9 branch of David Gibson's tree at https://github.com/dgibson/qemu.git
* It also relies on some work already posted here:
https://lists.gnu.org/archive/html/qemu-devel/2017-01/msg02527.html
Specifically patches 1..4 which set up the new CPU and MMU models.

Changes v1 -> v2:
Patch 1/12: scripts/update-linux-headers.sh: refactor extra files

I've factored the script to make it easier to add new files.

Patch 2/12: scripts/update-linux-headers.sh: add new files for ARM

* Added the two new arm headers.

Patch 3/12: Move virtio_mmio.h to fix update-linux-headers.sh
* FWIW, here's one way of fixing it.

Patch 4/12: Update headers using update-linux-headers.sh

* Added information about where the headers came from.

Patch 5/12: spapr: Add ibm,processor-radix-AP-encodings to the device tree

* ppc_radix_page_info now kept in native format, conversion to BE done when adding to the device tree.
* radix_page_info moved into the CPU class, cleaning up some code.

Patch 6/12: target-ppc: support KVM_CAP_PPC_MMU_RADIX, KVM_CAP_PPC_MMU_HASH_V3

* cap_mmu_hash renamed to cap_mmu_hash_v3.

Patch 7/12: spapr: Only setup HTP if necessary.

* This patch has been mostly rewritten to move the late HPT allocation to CAS. This allows a guest to start in radix mode (when it's in real mode) and then change to hash, even if it is a legacy guest and will not call h_register_process_table().
* Added an exported function to spapr.c to perform HPT allocation and adjust the vrma if necessary. This makes it possible to allocate the HPT from h_client_architecture_support() in spapr_hcall.c.

Patch 8/12: spapr: Add h_register_process_table() hypercall

* I haven't addressed review comments for this patch because it overlaps with Suraj's implementation of the same function and we'll work together to integrate them.

Patch 10/12: spapr: Enable ISA 3.0 MMU mode selection via CAS

* Unused bits removed.
* Logic and bit definitions changed due to architectural change.
* Cleanly terminate QEMU if the guest requests an unavailable mode (as required by the new architecture).
* Legacy guest workaround moved to it's own patch.
* I'm sorry for the bitfield constants in spapr_dt_ov5_platform_support() but there don't seem to be convienent macros for converting an option vector specifier (OV_BIT(x,y)) into a byte-mask. I'm open to suggestions.

Sam Bobroff (12):
spapr: Small cleanup of PPC MMU enums scripts/update-linux-headers.sh: refactor extra files scripts/update-linux-headers.sh: add new files for ARM Move virtio_mmio.h to fix update-linux-headers.sh Update headers using update-linux-headers.sh
spapr: Add ibm,processor-radix-AP-encodings to the device tree
target-ppc: support KVM_CAP_PPC_MMU_RADIX, KVM_CAP_PPC_MMU_HASH_V3
spapr: Only setup HTP if necessary.
spapr: Add h_register_process_table() hypercall
spapr: move spapr_populate_pa_features()
spapr: Enable ISA 3.0 MMU mode selection via CAS
spapr: Workaround for broken radix guests

 hw/ppc/spapr.c                                     | 190 +++++++---
 hw/ppc/spapr_hcall.c                               |  84 ++++-
 hw/virtio/virtio-mmio.c                            |   2 +-
 include/hw/ppc/spapr.h                             |   3 +
 include/hw/ppc/spapr_ovec.h                        |   8 +
 .../linux/virtio_mmio.h                            |   0
 include/standard-headers/linux/input-event-codes.h |   2 +-
 include/standard-headers/linux/pci_regs.h          |   8 +
 include/standard-headers/linux/virtio_ids.h        |   1 +
 include/sysemu/kvm.h                               |   1 +
 linux-headers/asm-arm/kvm.h                        |   2 +
 linux-headers/asm-arm/unistd-eabi.h                |   5 +
 linux-headers/asm-arm/unistd-oabi.h                |  17 +
 linux-headers/asm-arm/unistd.h                     | 419 +--------------------
 linux-headers/asm-powerpc/kvm.h                    |  27 ++
 linux-headers/asm-powerpc/unistd.h                 |   1 +
 linux-headers/asm-x86/kvm_para.h                   |   4 +-
 linux-headers/linux/kvm.h                          |  20 +-
 linux-headers/linux/vfio.h                         |  10 +
 scripts/update-linux-headers.sh                    |  26 +-
 target/ppc/cpu-qom.h                               |  13 +-
 target/ppc/cpu.h                                   |   4 +
 target/ppc/kvm.c                                   |  61 ++-
 target/ppc/kvm_ppc.h                               |  13 +
 target/ppc/mmu-hash64.c                            |  10 +-
 target/ppc/mmu_helper.c                            |  67 ++--
 target/ppc/translate.c                             |  12 +-
 27 files changed, 457 insertions(+), 553 deletions(-)
 rename include/{standard-headers => kernel-headers}/linux/virtio_mmio.h (100%)
 create mode 100644 linux-headers/asm-arm/unistd-eabi.h
 create mode 100644 linux-headers/asm-arm/unistd-oabi.h
Signature · 1 line
2.11.0
#1 lore raw

RFC PATCH v2 1/12 spapr: Small cleanup of PPC MMU enums

The PPC MMU types are sometimes treated as if they were a bit field and sometime as if they were an enum which causes maintenance
problems: flipping bits in the MMU type (which is done on both the 1TB segment and 64K segment bits) currently produces new MMU type values that are not handled in every "switch" on it, sometimes causing an abort().

This patch provides some macros that can be used to filter out the "bit field-like" bits so that the remainder of the value can be switched on, like an enum. This allows removal of all of the "degraded" types from the list and should ease maintenance.

---
 hw/ppc/spapr.c          |  8 +++---
 target/ppc/cpu-qom.h    | 12 ++++-----
 target/ppc/kvm.c        |  8 +++---
 target/ppc/mmu-hash64.c | 10 ++++----
 target/ppc/mmu_helper.c | 67 ++++++++++++++++++++-----------------------------
 target/ppc/translate.c  | 12 ++++-----
 6 files changed, 50 insertions(+), 67 deletions(-)
hw/ppc/spapr.c +3 -5
index 5904e6498f..cceb35f083 100644
@@ -359,14 +359,12 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
     uint8_t *pa_features;
     size_t pa_size;
 
-    switch (env->mmu_model) {
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
+    case POWERPC_MMU_VER_2_06:
         pa_features = pa_features_206;
         pa_size = sizeof(pa_features_206);
         break;
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
+    case POWERPC_MMU_VER_2_07:
         pa_features = pa_features_207;
         pa_size = sizeof(pa_features_207);
         break;
target/ppc/cpu-qom.h +6 -6
index 4e3132b56b..4807f4d86c 100644
@@ -79,21 +79,21 @@ enum powerpc_mmu_t {
     POWERPC_MMU_2_06       = POWERPC_MMU_64 | POWERPC_MMU_1TSEG
                              | POWERPC_MMU_64K
                              | POWERPC_MMU_AMR | 0x00000003,
-    /* Architecture 2.06 "degraded" (no 1T segments)           */
-    POWERPC_MMU_2_06a      = POWERPC_MMU_64 | POWERPC_MMU_AMR
-                             | 0x00000003,
     /* Architecture 2.07 variant                               */
     POWERPC_MMU_2_07       = POWERPC_MMU_64 | POWERPC_MMU_1TSEG
                              | POWERPC_MMU_64K
                              | POWERPC_MMU_AMR | 0x00000004,
-    /* Architecture 2.07 "degraded" (no 1T segments)           */
-    POWERPC_MMU_2_07a      = POWERPC_MMU_64 | POWERPC_MMU_AMR
-                             | 0x00000004,
     /* Architecture 3.00 variant                               */
     POWERPC_MMU_3_00       = POWERPC_MMU_64 | POWERPC_MMU_1TSEG
                              | POWERPC_MMU_64K
                              | POWERPC_MMU_AMR | 0x00000005,
 };
+#define POWERPC_MMU_VER(x) ((x) & (POWERPC_MMU_64 | 0xFFFF))
+#define POWERPC_MMU_VER_64B POWERPC_MMU_VER(POWERPC_MMU_64B)
+#define POWERPC_MMU_VER_2_03 POWERPC_MMU_VER(POWERPC_MMU_2_03)
+#define POWERPC_MMU_VER_2_06 POWERPC_MMU_VER(POWERPC_MMU_2_06)
+#define POWERPC_MMU_VER_2_07 POWERPC_MMU_VER(POWERPC_MMU_2_07)
+#define POWERPC_MMU_VER_3_00 POWERPC_MMU_VER(POWERPC_MMU_3_00)
 
 /*****************************************************************************/
 /* Exception model                                                           */
target/ppc/kvm.c +4 -4
index 52bbea514a..d53ede8b4a 100644
@@ -282,8 +282,8 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
             info->flags |= KVM_PPC_1T_SEGMENTS;
         }
 
-        if (env->mmu_model == POWERPC_MMU_2_06 ||
-            env->mmu_model == POWERPC_MMU_2_07) {
+        if (POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_06 ||
+           POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_07) {
             info->slb_size = 32;
         } else {
             info->slb_size = 64;
@@ -297,8 +297,8 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
         i++;
 
         /* 64K on MMU 2.06 and later */
-        if (env->mmu_model == POWERPC_MMU_2_06 ||
-            env->mmu_model == POWERPC_MMU_2_07) {
+        if (POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_06 ||
+            POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_07) {
             info->sps[i].page_shift = 16;
             info->sps[i].slb_enc = 0x110;
             info->sps[i].enc[0].page_shift = 16;
target/ppc/mmu-hash64.c +5 -5
index 76669ed82c..6346167b48 100644
@@ -1032,8 +1032,8 @@ void helper_store_lpcr(CPUPPCState *env, target_ulong val)
     uint64_t lpcr = 0;
 
     /* Filter out bits */
-    switch (env->mmu_model) {
-    case POWERPC_MMU_64B: /* 970 */
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
+    case POWERPC_MMU_VER_64B: /* 970 */
         if (val & 0x40) {
             lpcr |= LPCR_LPES0;
         }
@@ -1059,19 +1059,19 @@ void helper_store_lpcr(CPUPPCState *env, target_ulong val)
          * to dig HRMOR out of HID5
          */
         break;
-    case POWERPC_MMU_2_03: /* P5p */
+    case POWERPC_MMU_VER_2_03: /* P5p */
         lpcr = val & (LPCR_RMLS | LPCR_ILE |
                       LPCR_LPES0 | LPCR_LPES1 |
                       LPCR_RMI | LPCR_HDICE);
         break;
-    case POWERPC_MMU_2_06: /* P7 */
+    case POWERPC_MMU_VER_2_06: /* P7 */
         lpcr = val & (LPCR_VPM0 | LPCR_VPM1 | LPCR_ISL | LPCR_DPFD |
                       LPCR_VRMASD | LPCR_RMLS | LPCR_ILE |
                       LPCR_P7_PECE0 | LPCR_P7_PECE1 | LPCR_P7_PECE2 |
                       LPCR_MER | LPCR_TC |
                       LPCR_LPES0 | LPCR_LPES1 | LPCR_HDICE);
         break;
-    case POWERPC_MMU_2_07: /* P8 */
+    case POWERPC_MMU_VER_2_07: /* P8 */
         lpcr = val & (LPCR_VPM0 | LPCR_VPM1 | LPCR_ISL | LPCR_KBV |
                       LPCR_DPFD | LPCR_VRMASD | LPCR_RMLS | LPCR_ILE |
                       LPCR_AIL | LPCR_ONL | LPCR_P8_PECE0 | LPCR_P8_PECE1 |
target/ppc/mmu_helper.c +27 -40

115 lines not on this page · raw message

target/ppc/translate.c +5 -7
index b09e16ff76..2a24d1de67 100644
@@ -6988,18 +6988,16 @@ void ppc_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
     if (env->spr_cb[SPR_LPCR].name)
         cpu_fprintf(f, " LPCR " TARGET_FMT_lx "\n", env->spr[SPR_LPCR]);
 
-    switch (env->mmu_model) {
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
     case POWERPC_MMU_32B:
     case POWERPC_MMU_601:
     case POWERPC_MMU_SOFT_6xx:
     case POWERPC_MMU_SOFT_74xx:
 #if defined(TARGET_PPC64)
-    case POWERPC_MMU_64B:
-    case POWERPC_MMU_2_03:
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
+    case POWERPC_MMU_VER_64B:
+    case POWERPC_MMU_VER_2_03:
+    case POWERPC_MMU_VER_2_06:
+    case POWERPC_MMU_VER_2_07:
 #endif
         cpu_fprintf(f, " SDR1 " TARGET_FMT_lx "   DAR " TARGET_FMT_lx
                        "  DSISR " TARGET_FMT_lx "\n", env->spr[SPR_SDR1],
Signature · 1 line
2.11.0
#2 lore raw

Re: [Qemu-devel] [RFC PATCH v2 01/12] spapr: Small cleanup of PPC MMU enums

Quoted 7 lines · Sam Bobroff

The PPC MMU types are sometimes treated as if they were a bit field and sometime as if they were an enum which causes maintenance
problems: flipping bits in the MMU type (which is done on both the 1TB segment and 64K segment bits) currently produces new MMU type values that are not handled in every "switch" on it, sometimes causing an abort().

Quoted 6 lines

This patch provides some macros that can be used to filter out the "bit field-like" bits so that the remainder of the value can be switched on, like an enum. This allows removal of all of the "degraded" types from the list and should ease maintenance.

Seems like a good idea.

Quoted 281 lines
---
 hw/ppc/spapr.c          |  8 +++---
 target/ppc/cpu-qom.h    | 12 ++++-----
 target/ppc/kvm.c        |  8 +++---
 target/ppc/mmu-hash64.c | 10 ++++----
 target/ppc/mmu_helper.c | 67 ++++++++++++++++++++-----------------------------
 target/ppc/translate.c  | 12 ++++-----
 6 files changed, 50 insertions(+), 67 deletions(-)
hw/ppc/spapr.c +3 -5
index 5904e6498f..cceb35f083 100644
@@ -359,14 +359,12 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
     uint8_t *pa_features;
     size_t pa_size;
 
-    switch (env->mmu_model) {
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
+    case POWERPC_MMU_VER_2_06:
         pa_features = pa_features_206;
         pa_size = sizeof(pa_features_206);
         break;
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
+    case POWERPC_MMU_VER_2_07:
         pa_features = pa_features_207;
         pa_size = sizeof(pa_features_207);
         break;
target/ppc/cpu-qom.h +6 -6
index 4e3132b56b..4807f4d86c 100644
@@ -79,21 +79,21 @@ enum powerpc_mmu_t {
     POWERPC_MMU_2_06       = POWERPC_MMU_64 | POWERPC_MMU_1TSEG
                              | POWERPC_MMU_64K
                              | POWERPC_MMU_AMR | 0x00000003,
-    /* Architecture 2.06 "degraded" (no 1T segments)           */
-    POWERPC_MMU_2_06a      = POWERPC_MMU_64 | POWERPC_MMU_AMR
-                             | 0x00000003,
     /* Architecture 2.07 variant                               */
     POWERPC_MMU_2_07       = POWERPC_MMU_64 | POWERPC_MMU_1TSEG
                              | POWERPC_MMU_64K
                              | POWERPC_MMU_AMR | 0x00000004,
-    /* Architecture 2.07 "degraded" (no 1T segments)           */
-    POWERPC_MMU_2_07a      = POWERPC_MMU_64 | POWERPC_MMU_AMR
-                             | 0x00000004,
     /* Architecture 3.00 variant                               */
     POWERPC_MMU_3_00       = POWERPC_MMU_64 | POWERPC_MMU_1TSEG
                              | POWERPC_MMU_64K
                              | POWERPC_MMU_AMR | 0x00000005,
 };
+#define POWERPC_MMU_VER(x) ((x) & (POWERPC_MMU_64 | 0xFFFF))
+#define POWERPC_MMU_VER_64B POWERPC_MMU_VER(POWERPC_MMU_64B)
+#define POWERPC_MMU_VER_2_03 POWERPC_MMU_VER(POWERPC_MMU_2_03)
+#define POWERPC_MMU_VER_2_06 POWERPC_MMU_VER(POWERPC_MMU_2_06)
+#define POWERPC_MMU_VER_2_07 POWERPC_MMU_VER(POWERPC_MMU_2_07)
+#define POWERPC_MMU_VER_3_00 POWERPC_MMU_VER(POWERPC_MMU_3_00)
 
 /*****************************************************************************/
 /* Exception model                                                           */
target/ppc/kvm.c +4 -4
index 52bbea514a..d53ede8b4a 100644
@@ -282,8 +282,8 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
             info->flags |= KVM_PPC_1T_SEGMENTS;
         }
 
-        if (env->mmu_model == POWERPC_MMU_2_06 ||
-            env->mmu_model == POWERPC_MMU_2_07) {
+        if (POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_06 ||
+           POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_07) {
             info->slb_size = 32;
         } else {
             info->slb_size = 64;
@@ -297,8 +297,8 @@ static void kvm_get_fallback_smmu_info(PowerPCCPU *cpu,
         i++;
 
         /* 64K on MMU 2.06 and later */
-        if (env->mmu_model == POWERPC_MMU_2_06 ||
-            env->mmu_model == POWERPC_MMU_2_07) {
+        if (POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_06 ||
+            POWERPC_MMU_VER(env->mmu_model) == POWERPC_MMU_VER_2_07) {
             info->sps[i].page_shift = 16;
             info->sps[i].slb_enc = 0x110;
             info->sps[i].enc[0].page_shift = 16;
target/ppc/mmu-hash64.c +5 -5
index 76669ed82c..6346167b48 100644
@@ -1032,8 +1032,8 @@ void helper_store_lpcr(CPUPPCState *env, target_ulong val)
     uint64_t lpcr = 0;
 
     /* Filter out bits */
-    switch (env->mmu_model) {
-    case POWERPC_MMU_64B: /* 970 */
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
+    case POWERPC_MMU_VER_64B: /* 970 */
         if (val & 0x40) {
             lpcr |= LPCR_LPES0;
         }
@@ -1059,19 +1059,19 @@ void helper_store_lpcr(CPUPPCState *env, target_ulong val)
          * to dig HRMOR out of HID5
          */
         break;
-    case POWERPC_MMU_2_03: /* P5p */
+    case POWERPC_MMU_VER_2_03: /* P5p */
         lpcr = val & (LPCR_RMLS | LPCR_ILE |
                       LPCR_LPES0 | LPCR_LPES1 |
                       LPCR_RMI | LPCR_HDICE);
         break;
-    case POWERPC_MMU_2_06: /* P7 */
+    case POWERPC_MMU_VER_2_06: /* P7 */
         lpcr = val & (LPCR_VPM0 | LPCR_VPM1 | LPCR_ISL | LPCR_DPFD |
                       LPCR_VRMASD | LPCR_RMLS | LPCR_ILE |
                       LPCR_P7_PECE0 | LPCR_P7_PECE1 | LPCR_P7_PECE2 |
                       LPCR_MER | LPCR_TC |
                       LPCR_LPES0 | LPCR_LPES1 | LPCR_HDICE);
         break;
-    case POWERPC_MMU_2_07: /* P8 */
+    case POWERPC_MMU_VER_2_07: /* P8 */
         lpcr = val & (LPCR_VPM0 | LPCR_VPM1 | LPCR_ISL | LPCR_KBV |
                       LPCR_DPFD | LPCR_VRMASD | LPCR_RMLS | LPCR_ILE |
                       LPCR_AIL | LPCR_ONL | LPCR_P8_PECE0 | LPCR_P8_PECE1 |
target/ppc/mmu_helper.c +27 -40
index eb2d482ef7..0f6016ff0d 100644
@@ -1260,7 +1260,7 @@ static void mmu6xx_dump_mmu(FILE *f, fprintf_function cpu_fprintf,
 
 void dump_mmu(FILE *f, fprintf_function cpu_fprintf, CPUPPCState *env)
 {
-    switch (env->mmu_model) {
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
     case POWERPC_MMU_BOOKE:
         mmubooke_dump_mmu(f, cpu_fprintf, env);
         break;
@@ -1272,12 +1272,10 @@ void dump_mmu(FILE *f, fprintf_function cpu_fprintf, CPUPPCState *env)
         mmu6xx_dump_mmu(f, cpu_fprintf, env);
         break;
 #if defined(TARGET_PPC64)
-    case POWERPC_MMU_64B:
-    case POWERPC_MMU_2_03:
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
+    case POWERPC_MMU_VER_64B:
+    case POWERPC_MMU_VER_2_03:
+    case POWERPC_MMU_VER_2_06:
+    case POWERPC_MMU_VER_2_07:
         dump_slb(f, cpu_fprintf, ppc_env_get_cpu(env));
         break;
 #endif
@@ -1412,14 +1410,12 @@ hwaddr ppc_cpu_get_phys_page_debug(CPUState *cs, vaddr addr)
     CPUPPCState *env = &cpu->env;
     mmu_ctx_t ctx;
 
-    switch (env->mmu_model) {
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
 #if defined(TARGET_PPC64)
-    case POWERPC_MMU_64B:
-    case POWERPC_MMU_2_03:
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
+    case POWERPC_MMU_VER_64B:
+    case POWERPC_MMU_VER_2_03:
+    case POWERPC_MMU_VER_2_06:
+    case POWERPC_MMU_VER_2_07:
         return ppc_hash64_get_phys_page_debug(cpu, addr);
 #endif
 
@@ -1904,6 +1900,12 @@ void ppc_tlb_invalidate_all(CPUPPCState *env)
 {
     PowerPCCPU *cpu = ppc_env_get_cpu(env);
 
+#if defined(TARGET_PPC64)
+    if (env->mmu_model & POWERPC_MMU_64) {
+        env->tlb_need_flush = 0;
+        tlb_flush(CPU(cpu));
+    } else
+#endif /* defined(TARGET_PPC64) */
     switch (env->mmu_model) {
     case POWERPC_MMU_SOFT_6xx:
     case POWERPC_MMU_SOFT_74xx:
@@ -1928,21 +1930,12 @@ void ppc_tlb_invalidate_all(CPUPPCState *env)
         break;
     case POWERPC_MMU_32B:
     case POWERPC_MMU_601:
-#if defined(TARGET_PPC64)
-    case POWERPC_MMU_64B:
-    case POWERPC_MMU_2_03:
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
-    case POWERPC_MMU_3_00:
-#endif /* defined(TARGET_PPC64) */
         env->tlb_need_flush = 0;
         tlb_flush(CPU(cpu));
         break;
     default:
         /* XXX: TODO */
-        cpu_abort(CPU(cpu), "Unknown MMU model %d\n", env->mmu_model);
+        cpu_abort(CPU(cpu), "Unknown MMU model %x\n", env->mmu_model);
         break;
     }
 }
@@ -1951,6 +1944,16 @@ void ppc_tlb_invalidate_one(CPUPPCState *env, target_ulong addr)
 {
 #if !defined(FLUSH_ALL_TLBS)
     addr &= TARGET_PAGE_MASK;
+#if defined(TARGET_PPC64)
+    if (env->mmu_model & POWERPC_MMU_64) {
+        /* tlbie invalidate TLBs for all segments */
+        /* XXX: given the fact that there are too many segments to invalidate,
+         *      and we still don't have a tlb_flush_mask(env, n, mask) in QEMU,
+         *      we just invalidate all TLBs
+         */
+        env->tlb_need_flush |= TLB_NEED_LOCAL_FLUSH;
+    } else
+#endif /* defined(TARGET_PPC64) */
     switch (env->mmu_model) {
     case POWERPC_MMU_SOFT_6xx:
     case POWERPC_MMU_SOFT_74xx:
@@ -1968,22 +1971,6 @@ void ppc_tlb_invalidate_one(CPUPPCState *env, target_ulong addr)
          */
         env->tlb_need_flush |= TLB_NEED_LOCAL_FLUSH;
         break;
-#if defined(TARGET_PPC64)
-    case POWERPC_MMU_64B:
-    case POWERPC_MMU_2_03:
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
-    case POWERPC_MMU_3_00:
-        /* tlbie invalidate TLBs for all segments */
-        /* XXX: given the fact that there are too many segments to invalidate,
-         *      and we still don't have a tlb_flush_mask(env, n, mask) in QEMU,
-         *      we just invalidate all TLBs
-         */
-        env->tlb_need_flush |= TLB_NEED_LOCAL_FLUSH;
-        break;
-#endif /* defined(TARGET_PPC64) */
     default:
         /* Should never reach here with other MMU models */
         assert(0);
target/ppc/translate.c +5 -7
index b09e16ff76..2a24d1de67 100644
@@ -6988,18 +6988,16 @@ void ppc_cpu_dump_state(CPUState *cs, FILE *f, fprintf_function cpu_fprintf,
     if (env->spr_cb[SPR_LPCR].name)
         cpu_fprintf(f, " LPCR " TARGET_FMT_lx "\n", env->spr[SPR_LPCR]);
 
-    switch (env->mmu_model) {
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
     case POWERPC_MMU_32B:
     case POWERPC_MMU_601:
     case POWERPC_MMU_SOFT_6xx:
     case POWERPC_MMU_SOFT_74xx:
 #if defined(TARGET_PPC64)
-    case POWERPC_MMU_64B:
-    case POWERPC_MMU_2_03:
-    case POWERPC_MMU_2_06:
-    case POWERPC_MMU_2_06a:
-    case POWERPC_MMU_2_07:
-    case POWERPC_MMU_2_07a:
+    case POWERPC_MMU_VER_64B:
+    case POWERPC_MMU_VER_2_03:
+    case POWERPC_MMU_VER_2_06:
+    case POWERPC_MMU_VER_2_07:
 #endif
         cpu_fprintf(f, " SDR1 " TARGET_FMT_lx "   DAR " TARGET_FMT_lx
                        "  DSISR " TARGET_FMT_lx "\n", env->spr[SPR_SDR1],
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
#3 lore raw

RFC PATCH v2 2/12 scripts/update-linux-headers.sh: refactor extra files

In reply to Sam Bobroff

Refactor the architecture specific code to make it easier to add new special case files.

There should be no change in functionality.

---
v2:

I've factored the script to make it easier to add new files.

 scripts/update-linux-headers.sh | 25 +++++++++++--------------
 1 file changed, 11 insertions(+), 14 deletions(-)
scripts/update-linux-headers.sh +11 -14
index ef11a8ab42..c75c30da1b 100755
@@ -76,28 +76,25 @@ for arch in $ARCHLIST; do
     fi
 
     make -C "$linux" INSTALL_HDR_PATH="$tmpdir" ARCH=$arch headers_install
+    ARCH_EXTRA=
+    ARCH_STD_EXTRA=
+    case "$arch" in
+        powerpc) ARCH_EXTRA=epapr_hcalls.h ;;
+        s390) ARCH_STD_EXTRA="kvm_virtio.h virtio-ccw.h" ;;
+        x86) ARCH_EXTRA="unistd_32.h unistd_x32.h unistd_64.h" ARCH_STD_EXTRA="hyperv.h" ;;
+    esac
 
     rm -rf "$output/linux-headers/asm-$arch"
     mkdir -p "$output/linux-headers/asm-$arch"
-    for header in kvm.h kvm_para.h unistd.h; do
+    for header in kvm.h kvm_para.h unistd.h $ARCH_EXTRA; do
         cp "$tmpdir/include/asm/$header" "$output/linux-headers/asm-$arch"
     done
-    if [ $arch = powerpc ]; then
-        cp "$tmpdir/include/asm/epapr_hcalls.h" "$output/linux-headers/asm-powerpc/"
-    fi
 
     rm -rf "$output/include/standard-headers/asm-$arch"
     mkdir -p "$output/include/standard-headers/asm-$arch"
-    if [ $arch = s390 ]; then
-        cp_portable "$tmpdir/include/asm/kvm_virtio.h" "$output/include/standard-headers/asm-s390/"
-        cp_portable "$tmpdir/include/asm/virtio-ccw.h" "$output/include/standard-headers/asm-s390/"
-    fi
-    if [ $arch = x86 ]; then
-        cp_portable "$tmpdir/include/asm/hyperv.h" "$output/include/standard-headers/asm-x86/"
-        cp "$tmpdir/include/asm/unistd_32.h" "$output/linux-headers/asm-x86/"
-        cp "$tmpdir/include/asm/unistd_x32.h" "$output/linux-headers/asm-x86/"
-        cp "$tmpdir/include/asm/unistd_64.h" "$output/linux-headers/asm-x86/"
-    fi
+    for header in $ARCH_STD_EXTRA; do
+        cp_portable "$tmpdir/include/asm/$header" "$output/include/standard-headers/asm-$arch/"
+    done
 done
 
 rm -rf "$output/linux-headers/linux"
Signature · 1 line
2.11.0
Reviewed-by on v2 2/12
#4 lore raw

Re: [Qemu-devel] [RFC PATCH v2 02/12] scripts/update-linux-headers.sh: refactor extra files

Quoted 6 lines · Sam Bobroff

Refactor the architecture specific code to make it easier to add new special case files.

There should be no change in functionality.

Quoted 52 lines

---
v2:

I've factored the script to make it easier to add new files.

 scripts/update-linux-headers.sh | 25 +++++++++++--------------
 1 file changed, 11 insertions(+), 14 deletions(-)
scripts/update-linux-headers.sh +11 -14
index ef11a8ab42..c75c30da1b 100755
@@ -76,28 +76,25 @@ for arch in $ARCHLIST; do
     fi
 
     make -C "$linux" INSTALL_HDR_PATH="$tmpdir" ARCH=$arch headers_install
+    ARCH_EXTRA=
+    ARCH_STD_EXTRA=
+    case "$arch" in
+        powerpc) ARCH_EXTRA=epapr_hcalls.h ;;
+        s390) ARCH_STD_EXTRA="kvm_virtio.h virtio-ccw.h" ;;
+        x86) ARCH_EXTRA="unistd_32.h unistd_x32.h unistd_64.h" ARCH_STD_EXTRA="hyperv.h" ;;
+    esac
 
     rm -rf "$output/linux-headers/asm-$arch"
     mkdir -p "$output/linux-headers/asm-$arch"
-    for header in kvm.h kvm_para.h unistd.h; do
+    for header in kvm.h kvm_para.h unistd.h $ARCH_EXTRA; do
         cp "$tmpdir/include/asm/$header" "$output/linux-headers/asm-$arch"
     done
-    if [ $arch = powerpc ]; then
-        cp "$tmpdir/include/asm/epapr_hcalls.h" "$output/linux-headers/asm-powerpc/"
-    fi
 
     rm -rf "$output/include/standard-headers/asm-$arch"
     mkdir -p "$output/include/standard-headers/asm-$arch"
-    if [ $arch = s390 ]; then
-        cp_portable "$tmpdir/include/asm/kvm_virtio.h" "$output/include/standard-headers/asm-s390/"
-        cp_portable "$tmpdir/include/asm/virtio-ccw.h" "$output/include/standard-headers/asm-s390/"
-    fi
-    if [ $arch = x86 ]; then
-        cp_portable "$tmpdir/include/asm/hyperv.h" "$output/include/standard-headers/asm-x86/"
-        cp "$tmpdir/include/asm/unistd_32.h" "$output/linux-headers/asm-x86/"
-        cp "$tmpdir/include/asm/unistd_x32.h" "$output/linux-headers/asm-x86/"
-        cp "$tmpdir/include/asm/unistd_64.h" "$output/linux-headers/asm-x86/"
-    fi
+    for header in $ARCH_STD_EXTRA; do
+        cp_portable "$tmpdir/include/asm/$header" "$output/include/standard-headers/asm-$arch/"
+    done
 done
 
 rm -rf "$output/linux-headers/linux"
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
#5 lore raw

RFC PATCH v2 3/12 scripts/update-linux-headers.sh: add new files for ARM

In reply to Sam Bobroff

The kernel has added some new headers for ARM, so add these so that the script can be run successfully.

---
v2:

* Added the two new arm headers.

 scripts/update-linux-headers.sh | 1 +
 1 file changed, 1 insertion(+)
scripts/update-linux-headers.sh +1 -0
index c75c30da1b..d8a178b070 100755
@@ -79,6 +79,7 @@ for arch in $ARCHLIST; do
     ARCH_EXTRA=
     ARCH_STD_EXTRA=
     case "$arch" in
+        arm) ARCH_EXTRA="unistd-eabi.h unistd-oabi.h" ;;
         powerpc) ARCH_EXTRA=epapr_hcalls.h ;;
         s390) ARCH_STD_EXTRA="kvm_virtio.h virtio-ccw.h" ;;
         x86) ARCH_EXTRA="unistd_32.h unistd_x32.h unistd_64.h" ARCH_STD_EXTRA="hyperv.h" ;;
Signature · 1 line
2.11.0
#6 lore raw

RFC PATCH v2 4/12 Move virtio_mmio.h to fix update-linux-headers.sh

In reply to Sam Bobroff

Currently, running update-linux-headers.sh will produce a patch that deletes virtio_mmio.h, which is still needed. This happens because virtio_mmio.h is in the directory used to store headers from the linux kernel that are copied by the kernel's "make headers_install" target (used by the update script) but it is not one of the files in that set.

Fix this by moving that file into a new directory.

In the future if that file is added to the "headers_install" target then this change should be reverted.

---
v2:
* FWIW, here's one way of fixing it.

 hw/virtio/virtio-mmio.c                                          | 2 +-
 include/{standard-headers => kernel-headers}/linux/virtio_mmio.h | 0
 2 files changed, 1 insertion(+), 1 deletion(-)
 rename include/{standard-headers => kernel-headers}/linux/virtio_mmio.h (100%)
hw/virtio/virtio-mmio.c +1 -1
index 5807aa87fe..cc6afa9da1 100644
@@ -20,7 +20,7 @@
  */
 
 #include "qemu/osdep.h"
-#include "standard-headers/linux/virtio_mmio.h"
+#include "kernel-headers/linux/virtio_mmio.h"
 #include "hw/sysbus.h"
 #include "hw/virtio/virtio.h"
 #include "qemu/host-utils.h"
include/kernel-headers/linux/virtio_mmio.h renamed from include/standard-headers/linux/virtio_mmio.h +0 -0
similarity index 100%
rename from include/standard-headers/linux/virtio_mmio.h
rename to include/kernel-headers/linux/virtio_mmio.h
Signature · 1 line
2.11.0
#7 lore raw

Re: [Qemu-devel] [RFC PATCH v2 04/12] Move virtio_mmio.h to fix update-linux-headers.sh

Quoted 7 lines · Sam Bobroff

Currently, running update-linux-headers.sh will produce a patch that deletes virtio_mmio.h, which is still needed. This happens because virtio_mmio.h is in the directory used to store headers from the linux kernel that are copied by the kernel's "make headers_install" target (used by the update script) but it is not one of the files in that set.

Quoted 6 lines

Fix this by moving that file into a new directory.

In the future if that file is added to the "headers_install" target then this change should be reverted.

This is a temporary condition. I'm merging a patch exporting virtio_mmio.h for next linux and we'll want to stay in sync, so I don't think we should make this change.

Quoted 29 lines

---
v2:
* FWIW, here's one way of fixing it.

 hw/virtio/virtio-mmio.c                                          | 2 +-
 include/{standard-headers => kernel-headers}/linux/virtio_mmio.h | 0
 2 files changed, 1 insertion(+), 1 deletion(-)
 rename include/{standard-headers => kernel-headers}/linux/virtio_mmio.h (100%)
hw/virtio/virtio-mmio.c +1 -1
index 5807aa87fe..cc6afa9da1 100644
@@ -20,7 +20,7 @@
  */
 
 #include "qemu/osdep.h"
-#include "standard-headers/linux/virtio_mmio.h"
+#include "kernel-headers/linux/virtio_mmio.h"
 #include "hw/sysbus.h"
 #include "hw/virtio/virtio.h"
 #include "qemu/host-utils.h"
include/kernel-headers/linux/virtio_mmio.h renamed from include/standard-headers/linux/virtio_mmio.h +0 -0
similarity index 100%
rename from include/standard-headers/linux/virtio_mmio.h
rename to include/kernel-headers/linux/virtio_mmio.h
Signature · 1 line
2.11.0
#8 lore raw

Re: [Qemu-devel] [RFC PATCH v2 04/12] Move virtio_mmio.h to fix update-linux-headers.sh

In reply to Sam Bobroff
Quoted 7 lines · Sam Bobroff

Currently, running update-linux-headers.sh will produce a patch that deletes virtio_mmio.h, which is still needed. This happens because virtio_mmio.h is in the directory used to store headers from the linux kernel that are copied by the kernel's "make headers_install" target (used by the update script) but it is not one of the files in that set.

Quoted 6 lines

Fix this by moving that file into a new directory.

In the future if that file is added to the "headers_install" target then this change should be reverted.

This is a temporary condition - I'm adding a patch exporting this header for next linux. So I don't think we should merge this.

Quoted 29 lines

---
v2:
* FWIW, here's one way of fixing it.

 hw/virtio/virtio-mmio.c                                          | 2 +-
 include/{standard-headers => kernel-headers}/linux/virtio_mmio.h | 0
 2 files changed, 1 insertion(+), 1 deletion(-)
 rename include/{standard-headers => kernel-headers}/linux/virtio_mmio.h (100%)
hw/virtio/virtio-mmio.c +1 -1
index 5807aa87fe..cc6afa9da1 100644
@@ -20,7 +20,7 @@
  */
 
 #include "qemu/osdep.h"
-#include "standard-headers/linux/virtio_mmio.h"
+#include "kernel-headers/linux/virtio_mmio.h"
 #include "hw/sysbus.h"
 #include "hw/virtio/virtio.h"
 #include "qemu/host-utils.h"
include/kernel-headers/linux/virtio_mmio.h renamed from include/standard-headers/linux/virtio_mmio.h +0 -0
similarity index 100%
rename from include/standard-headers/linux/virtio_mmio.h
rename to include/kernel-headers/linux/virtio_mmio.h
Signature · 1 line
2.11.0
#9 lore raw

Re: [Qemu-devel] [RFC PATCH v2 04/12] Move virtio_mmio.h to fix update-linux-headers.sh

Quoted 10 lines · Michael S. Tsirkin
Quoted 9 lines · Sam Bobroff

Currently, running update-linux-headers.sh will produce a patch that deletes virtio_mmio.h, which is still needed. This happens because virtio_mmio.h is in the directory used to store headers from the linux kernel that are copied by the kernel's "make headers_install" target (used by the update script) but it is not one of the files in that set.

Fix this by moving that file into a new directory.

Quoted 7 lines
Quoted 4 lines

In the future if that file is added to the "headers_install" target then this change should be reverted.

This is a temporary condition - I'm adding a patch exporting this header for next linux. So I don't think we should merge this.

Great :-)

I'll drop this patch from the next version.

Cheers,
Sam.

Quoted 29 lines
Quoted 29 lines

---
v2:
* FWIW, here's one way of fixing it.

 hw/virtio/virtio-mmio.c                                          | 2 +-
 include/{standard-headers => kernel-headers}/linux/virtio_mmio.h | 0
 2 files changed, 1 insertion(+), 1 deletion(-)
 rename include/{standard-headers => kernel-headers}/linux/virtio_mmio.h (100%)
hw/virtio/virtio-mmio.c +1 -1
index 5807aa87fe..cc6afa9da1 100644
@@ -20,7 +20,7 @@
  */
 
 #include "qemu/osdep.h"
-#include "standard-headers/linux/virtio_mmio.h"
+#include "kernel-headers/linux/virtio_mmio.h"
 #include "hw/sysbus.h"
 #include "hw/virtio/virtio.h"
 #include "qemu/host-utils.h"
include/kernel-headers/linux/virtio_mmio.h renamed from include/standard-headers/linux/virtio_mmio.h +0 -0
similarity index 100%
rename from include/standard-headers/linux/virtio_mmio.h
rename to include/kernel-headers/linux/virtio_mmio.h
Signature · 1 line
2.11.0
#10 lore raw

RFC PATCH v2 5/12 Update headers using update-linux-headers.sh

In reply to Sam Bobroff

Updated against Paul's kvm-ppc-next tree:
git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc.git
... at commit:
5982f0849e08fe4e4e7df5e345c4539ce9780b1b
... in order to provide some new definitions needed by ISA 3.00 guests.

This is a large change because it is the first import since some kernel header files have become autogenerated.

---
v2:

* Added information about where the headers came from.

 include/standard-headers/linux/input-event-codes.h |   2 +-
 include/standard-headers/linux/pci_regs.h          |   8 +
 include/standard-headers/linux/virtio_ids.h        |   1 +
 linux-headers/asm-arm/kvm.h                        |   2 +
 linux-headers/asm-arm/unistd-eabi.h                |   5 +
 linux-headers/asm-arm/unistd-oabi.h                |  17 +
 linux-headers/asm-arm/unistd.h                     | 419 +--------------------
 linux-headers/asm-powerpc/kvm.h                    |  27 ++
 linux-headers/asm-powerpc/unistd.h                 |   1 +
 linux-headers/asm-x86/kvm_para.h                   |   4 +-
 linux-headers/linux/kvm.h                          |  20 +-
 linux-headers/linux/vfio.h                         |  10 +
 12 files changed, 97 insertions(+), 419 deletions(-)
 create mode 100644 linux-headers/asm-arm/unistd-eabi.h
 create mode 100644 linux-headers/asm-arm/unistd-oabi.h
include/standard-headers/linux/input-event-codes.h +1 -1
index 5c10f7e25d..c8b3338375 100644
@@ -640,7 +640,7 @@
  * Control a data application associated with the currently viewed channel,
  * e.g. teletext or data broadcast application (MHEG, MHP, HbbTV, etc.)
  */
-#define KEY_DATA			0x275
+#define KEY_DATA			0x277
 
 #define BTN_TRIGGER_HAPPY		0x2c0
 #define BTN_TRIGGER_HAPPY1		0x2c0
include/standard-headers/linux/pci_regs.h +8 -0
index e5a2e68b22..174d114708 100644
@@ -23,6 +23,14 @@
 #define LINUX_PCI_REGS_H
 
 /*
+ * Conventional PCI and PCI-X Mode 1 devices have 256 bytes of
+ * configuration space.  PCI-X Mode 2 and PCIe devices have 4096 bytes of
+ * configuration space.
+ */
+#define PCI_CFG_SPACE_SIZE	256
+#define PCI_CFG_SPACE_EXP_SIZE	4096
+
+/*
  * Under PCI, each device has 256 bytes of configuration address space,
  * of which the first 64 bytes are standardized as follows:
  */
include/standard-headers/linux/virtio_ids.h +1 -0
index fe74e422d4..6d5c3b2d4f 100644
@@ -43,4 +43,5 @@
 #define VIRTIO_ID_INPUT        18 /* virtio input */
 #define VIRTIO_ID_VSOCK        19 /* virtio vsock transport */
 #define VIRTIO_ID_CRYPTO       20 /* virtio crypto */
+
 #endif /* _LINUX_VIRTIO_IDS_H */
linux-headers/asm-arm/kvm.h +2 -0
index 2fb7859465..09a555cc83 100644
@@ -87,9 +87,11 @@ struct kvm_regs {
 /* Supported VGICv3 address types  */
 #define KVM_VGIC_V3_ADDR_TYPE_DIST	2
 #define KVM_VGIC_V3_ADDR_TYPE_REDIST	3
+#define KVM_VGIC_ITS_ADDR_TYPE		4
 
 #define KVM_VGIC_V3_DIST_SIZE		SZ_64K
 #define KVM_VGIC_V3_REDIST_SIZE		(2 * SZ_64K)
+#define KVM_VGIC_V3_ITS_SIZE		(2 * SZ_64K)
 
 #define KVM_ARM_VCPU_POWER_OFF		0 /* CPU is started in OFF state */
 #define KVM_ARM_VCPU_PSCI_0_2		1 /* CPU uses PSCI v0.2 */
linux-headers/asm-arm/unistd-eabi.h new file +5 -0
new file mode 100644
index 0000000000..266f1fcdfb
@@ -0,0 +1,5 @@
+#ifndef _ASM_ARM_UNISTD_EABI_H
+#define _ASM_ARM_UNISTD_EABI_H 1
+
+
+#endif /* _ASM_ARM_UNISTD_EABI_H */
linux-headers/asm-arm/unistd-oabi.h new file +17 -0
new file mode 100644
index 0000000000..47d9afb96d
@@ -0,0 +1,17 @@
+#ifndef _ASM_ARM_UNISTD_OABI_H
+#define _ASM_ARM_UNISTD_OABI_H 1
+
+#define __NR_time (__NR_SYSCALL_BASE + 13)
+#define __NR_umount (__NR_SYSCALL_BASE + 22)
+#define __NR_stime (__NR_SYSCALL_BASE + 25)
+#define __NR_alarm (__NR_SYSCALL_BASE + 27)
+#define __NR_utime (__NR_SYSCALL_BASE + 30)
+#define __NR_getrlimit (__NR_SYSCALL_BASE + 76)
+#define __NR_select (__NR_SYSCALL_BASE + 82)
+#define __NR_readdir (__NR_SYSCALL_BASE + 89)
+#define __NR_mmap (__NR_SYSCALL_BASE + 90)
+#define __NR_socketcall (__NR_SYSCALL_BASE + 102)
+#define __NR_syscall (__NR_SYSCALL_BASE + 113)
+#define __NR_ipc (__NR_SYSCALL_BASE + 117)
+
+#endif /* _ASM_ARM_UNISTD_OABI_H */
linux-headers/asm-arm/unistd.h +3 -416

434 lines not on this page · raw message

linux-headers/asm-powerpc/kvm.h +27 -0
index c93cf35ce3..4edbe4bb0e 100644
@@ -413,6 +413,26 @@ struct kvm_get_htab_header {
 	__u16	n_invalid;
 };
 
+/* For KVM_PPC_CONFIGURE_V3_MMU */
+struct kvm_ppc_mmuv3_cfg {
+	__u64	flags;
+	__u64	process_table;	/* second doubleword of partition table entry */
+};
+
+/* Flag values for KVM_PPC_CONFIGURE_V3_MMU */
+#define KVM_PPC_MMUV3_RADIX	1	/* 1 = radix mode, 0 = HPT */
+#define KVM_PPC_MMUV3_GTSE	2	/* global translation shootdown enb. */
+
+/* For KVM_PPC_GET_RMMU_INFO */
+struct kvm_ppc_rmmu_info {
+	struct kvm_ppc_radix_geom {
+		__u8	page_shift;
+		__u8	level_bits[4];
+		__u8	pad[3];
+	}	geometries[8];
+	__u32	ap_encodings[8];
+};
+
 /* Per-vcpu XICS interrupt controller state */
 #define KVM_REG_PPC_ICP_STATE	(KVM_REG_PPC | KVM_REG_SIZE_U64 | 0x8c)
 
@@ -573,6 +593,10 @@ struct kvm_get_htab_header {
 #define KVM_REG_PPC_SPRG9	(KVM_REG_PPC | KVM_REG_SIZE_U64 | 0xba)
 #define KVM_REG_PPC_DBSR	(KVM_REG_PPC | KVM_REG_SIZE_U32 | 0xbb)
 
+/* POWER9 registers */
+#define KVM_REG_PPC_TIDR	(KVM_REG_PPC | KVM_REG_SIZE_U64 | 0xbc)
+#define KVM_REG_PPC_PSSCR	(KVM_REG_PPC | KVM_REG_SIZE_U64 | 0xbd)
+
 /* Transactional Memory checkpointed state:
  * This is all GPRs, all VSX regs and a subset of SPRs
  */
@@ -596,6 +620,7 @@ struct kvm_get_htab_header {
 #define KVM_REG_PPC_TM_VSCR	(KVM_REG_PPC_TM | KVM_REG_SIZE_U32 | 0x67)
 #define KVM_REG_PPC_TM_DSCR	(KVM_REG_PPC_TM | KVM_REG_SIZE_U64 | 0x68)
 #define KVM_REG_PPC_TM_TAR	(KVM_REG_PPC_TM | KVM_REG_SIZE_U64 | 0x69)
+#define KVM_REG_PPC_TM_XER	(KVM_REG_PPC_TM | KVM_REG_SIZE_U64 | 0x6a)
 
 /* PPC64 eXternal Interrupt Controller Specification */
 #define KVM_DEV_XICS_GRP_SOURCES	1	/* 64-bit source attributes */
@@ -608,5 +633,7 @@ struct kvm_get_htab_header {
 #define  KVM_XICS_LEVEL_SENSITIVE	(1ULL << 40)
 #define  KVM_XICS_MASKED		(1ULL << 41)
 #define  KVM_XICS_PENDING		(1ULL << 42)
+#define  KVM_XICS_PRESENTED		(1ULL << 43)
+#define  KVM_XICS_QUEUED		(1ULL << 44)
 
 #endif /* __LINUX_KVM_POWERPC_H */
linux-headers/asm-powerpc/unistd.h +1 -0
index 1e66eba4c6..598043c7b6 100644
@@ -392,5 +392,6 @@
 #define __NR_copy_file_range	379
 #define __NR_preadv2		380
 #define __NR_pwritev2		381
+#define __NR_kexec_file_load	382
 
 #endif /* _ASM_POWERPC_UNISTD_H_ */
linux-headers/asm-x86/kvm_para.h +3 -1
index e41c5c1a28..0739a74626 100644
@@ -45,7 +45,9 @@ struct kvm_steal_time {
 	__u64 steal;
 	__u32 version;
 	__u32 flags;
-	__u32 pad[12];
+	__u8  preempted;
+	__u8  u8_pad[3];
+	__u32 pad[11];
 };
 
 #define KVM_STEAL_ALIGNMENT_BITS 5
linux-headers/linux/kvm.h +19 -1
index bb0ed71223..8391bbd21b 100644
@@ -651,6 +651,9 @@ struct kvm_enable_cap {
 };
 
 /* for KVM_PPC_GET_PVINFO */
+
+#define KVM_PPC_PVINFO_FLAGS_EV_IDLE   (1<<0)
+
 struct kvm_ppc_pvinfo {
 	/* out */
 	__u32 flags;
@@ -682,7 +685,12 @@ struct kvm_ppc_smmu_info {
 	struct kvm_ppc_one_seg_page_size sps[KVM_PPC_PAGE_SIZES_MAX_SZ];
 };
 
-#define KVM_PPC_PVINFO_FLAGS_EV_IDLE   (1<<0)
+/* for KVM_PPC_RESIZE_HPT_{PREPARE,COMMIT} */
+struct kvm_ppc_resize_hpt {
+	__u64 flags;
+	__u32 shift;
+	__u32 pad;
+};
 
 #define KVMIO 0xAE
 
@@ -870,6 +878,9 @@ struct kvm_ppc_smmu_info {
 #define KVM_CAP_S390_USER_INSTR0 130
 #define KVM_CAP_MSI_DEVID 131
 #define KVM_CAP_PPC_HTM 132
+#define KVM_CAP_SPAPR_RESIZE_HPT 133
+#define KVM_CAP_PPC_MMU_RADIX 134
+#define KVM_CAP_PPC_MMU_HASH_V3 135
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
@@ -1186,6 +1197,13 @@ struct kvm_s390_ucas_mapping {
 #define KVM_ARM_SET_DEVICE_ADDR	  _IOW(KVMIO,  0xab, struct kvm_arm_device_addr)
 /* Available with KVM_CAP_PPC_RTAS */
 #define KVM_PPC_RTAS_DEFINE_TOKEN _IOW(KVMIO,  0xac, struct kvm_rtas_token_args)
+/* Available with KVM_CAP_SPAPR_RESIZE_HPT */
+#define KVM_PPC_RESIZE_HPT_PREPARE _IOR(KVMIO, 0xad, struct kvm_ppc_resize_hpt)
+#define KVM_PPC_RESIZE_HPT_COMMIT  _IOR(KVMIO, 0xae, struct kvm_ppc_resize_hpt)
+/* Available with KVM_CAP_PPC_RADIX_MMU or KVM_CAP_PPC_HASH_MMU_V3 */
+#define KVM_PPC_CONFIGURE_V3_MMU  _IOW(KVMIO,  0xaf, struct kvm_ppc_mmuv3_cfg)
+/* Available with KVM_CAP_PPC_RADIX_MMU */
+#define KVM_PPC_GET_RMMU_INFO	  _IOW(KVMIO,  0xb0, struct kvm_ppc_rmmu_info)
 
 /* ioctl for vm fd */
 #define KVM_CREATE_DEVICE	  _IOWR(KVMIO,  0xe0, struct kvm_create_device)
linux-headers/linux/vfio.h +10 -0
index 759b850a3e..531cb2eda9 100644
@@ -203,6 +203,16 @@ struct vfio_device_info {
 };
 #define VFIO_DEVICE_GET_INFO		_IO(VFIO_TYPE, VFIO_BASE + 7)
 
+/*
+ * Vendor driver using Mediated device framework should provide device_api
+ * attribute in supported type attribute groups. Device API string should be one
+ * of the following corresponding to device flags in vfio_device_info structure.
+ */
+
+#define VFIO_DEVICE_API_PCI_STRING		"vfio-pci"
+#define VFIO_DEVICE_API_PLATFORM_STRING		"vfio-platform"
+#define VFIO_DEVICE_API_AMBA_STRING		"vfio-amba"
+
 /**
  * VFIO_DEVICE_GET_REGION_INFO - _IOWR(VFIO_TYPE, VFIO_BASE + 8,
  *				       struct vfio_region_info)
Signature · 1 line
2.11.0
#11 lore raw

RFC PATCH v2 6/12 spapr: Add ibm, processor-radix-AP-encodings to the device tree

In reply to Sam Bobroff

Use the new ioctl, KVM_PPC_GET_RMMU_INFO, to fetch radix MMU information from KVM and present the page encodings in the device tree under ibm,processor-radix-AP-encodings. This provides page size information to the guest which is necessary for it to use radix mode.

---
v2:

* ppc_radix_page_info now kept in native format, conversion to BE done when adding to the device tree.
* radix_page_info moved into the CPU class, cleaning up some code.

 hw/ppc/spapr.c       | 12 ++++++++++++
 include/sysemu/kvm.h |  1 +
 target/ppc/cpu-qom.h |  1 +
 target/ppc/cpu.h     |  4 ++++
 target/ppc/kvm.c     | 27 +++++++++++++++++++++++++++
 5 files changed, 45 insertions(+)
hw/ppc/spapr.c +12 -0
index cceb35f083..ca3812555f 100644
@@ -409,6 +409,8 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
     sPAPRDRConnector *drc;
     sPAPRDRConnectorClass *drck;
     int drc_index;
+    uint32_t radix_AP_encodings[PPC_PAGE_SIZES_MAX_SZ];
+    int i;
 
     drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index);
     if (drc) {
@@ -494,6 +496,16 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
 
     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt));
+
+    if (pcc->radix_page_info) {
+        for (i = 0; i < pcc->radix_page_info->count; i++) {
+            radix_AP_encodings[i] = cpu_to_be32(pcc->radix_page_info->entries[i]);
+        }
+        _FDT((fdt_setprop(fdt, offset, "ibm,processor-radix-AP-encodings",
+                          radix_AP_encodings,
+                          pcc->radix_page_info->count *
+                          sizeof(radix_AP_encodings[0]))));
+    }
 }
 
 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
include/sysemu/kvm.h +1 -0
index 3045ee7678..01a8db1180 100644
@@ -526,5 +526,6 @@ int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
  * Returns: 0 on success, or a negative errno on failure.
  */
 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
+struct ppc_radix_page_info *kvm_get_radix_page_info(void);
 int kvm_get_max_memslots(void);
 #endif
target/ppc/cpu-qom.h +1 -0
index 4807f4d86c..0efb543912 100644
@@ -195,6 +195,7 @@ typedef struct PowerPCCPUClass {
     int bfd_mach;
     uint32_t l1_dcache_size, l1_icache_size;
     const struct ppc_segment_page_sizes *sps;
+    struct ppc_radix_page_info *radix_page_info;
     void (*init_proc)(CPUPPCState *env);
     int  (*check_pow)(CPUPPCState *env);
     int (*handle_mmu_fault)(PowerPCCPU *cpu, vaddr eaddr, int rwx, int mmu_idx);
target/ppc/cpu.h +4 -0
index b559b67073..a6c8c5ff4c 100644
@@ -934,6 +934,10 @@ struct ppc_segment_page_sizes {
     struct ppc_one_seg_page_size sps[PPC_PAGE_SIZES_MAX_SZ];
 };
 
+struct ppc_radix_page_info {
+    uint32_t count;
+    uint32_t entries[PPC_PAGE_SIZES_MAX_SZ];
+};
 
 /*****************************************************************************/
 /* The whole PowerPC CPU context */
target/ppc/kvm.c +27 -0
index d53ede8b4a..cf62a42c1f 100644
@@ -48,6 +48,7 @@
 #if defined(TARGET_PPC64)
 #include "hw/ppc/spapr_cpu_core.h"
 #endif
+#include "sysemu/kvm_int.h"
 
 //#define DEBUG_KVM
 
@@ -329,6 +330,30 @@ static void kvm_get_smmu_info(PowerPCCPU *cpu, struct kvm_ppc_smmu_info *info)
     kvm_get_fallback_smmu_info(cpu, info);
 }
 
+struct ppc_radix_page_info *kvm_get_radix_page_info(void)
+{
+    KVMState *s = KVM_STATE(current_machine->accelerator);
+    struct ppc_radix_page_info *radix_page_info;
+    struct kvm_ppc_rmmu_info rmmu_info;
+    int i;
+
+    if (!kvm_check_extension(s, KVM_CAP_PPC_MMU_RADIX)) {
+        return NULL;
+    }
+    if (kvm_vm_ioctl(s, KVM_PPC_GET_RMMU_INFO, &rmmu_info)) {
+        return NULL;
+    }
+    radix_page_info = g_malloc0(sizeof(*radix_page_info));
+    radix_page_info->count = 0;
+    for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
+        if (rmmu_info.ap_encodings[i]) {
+            radix_page_info->entries[i] = rmmu_info.ap_encodings[i];
+            radix_page_info->count++;
+        }
+    }
+    return radix_page_info;
+}
+
 static long gethugepagesize(const char *mem_path)
 {
     struct statfs fs;
@@ -2379,6 +2404,8 @@ static void kvmppc_host_cpu_class_init(ObjectClass *oc, void *data)
         pcc->l1_icache_size = icache_size;
     }
 
+    pcc->radix_page_info = kvm_enabled() ? kvm_get_radix_page_info() : NULL;
+
     /* Reason: kvmppc_host_cpu_initfn() dies when !kvm_enabled() */
     dc->cannot_destroy_with_object_finalize_yet = true;
 }
Signature · 1 line
2.11.0
#12 lore raw

Re: [Qemu-devel] [RFC PATCH v2 06/12] spapr: Add ibm, processor-radix-AP-encodings to the device tree

Quoted 11 lines · Sam Bobroff

Use the new ioctl, KVM_PPC_GET_RMMU_INFO, to fetch radix MMU information from KVM and present the page encodings in the device tree under ibm,processor-radix-AP-encodings. This provides page size information to the guest which is necessary for it to use radix mode.

---
v2:

* ppc_radix_page_info now kept in native format, conversion to BE done when adding to the device tree.
* radix_page_info moved into the CPU class, cleaning up some code.

Looks pretty good, although I imaginge it will need a little rework to rebase on top of the TCG radix stuff. Also one comment below..

Quoted 120 lines
 hw/ppc/spapr.c       | 12 ++++++++++++
 include/sysemu/kvm.h |  1 +
 target/ppc/cpu-qom.h |  1 +
 target/ppc/cpu.h     |  4 ++++
 target/ppc/kvm.c     | 27 +++++++++++++++++++++++++++
 5 files changed, 45 insertions(+)
hw/ppc/spapr.c +12 -0
index cceb35f083..ca3812555f 100644
@@ -409,6 +409,8 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
     sPAPRDRConnector *drc;
     sPAPRDRConnectorClass *drck;
     int drc_index;
+    uint32_t radix_AP_encodings[PPC_PAGE_SIZES_MAX_SZ];
+    int i;
 
     drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index);
     if (drc) {
@@ -494,6 +496,16 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
     _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));
 
     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt));
+
+    if (pcc->radix_page_info) {
+        for (i = 0; i < pcc->radix_page_info->count; i++) {
+            radix_AP_encodings[i] = cpu_to_be32(pcc->radix_page_info->entries[i]);
+        }
+        _FDT((fdt_setprop(fdt, offset, "ibm,processor-radix-AP-encodings",
+                          radix_AP_encodings,
+                          pcc->radix_page_info->count *
+                          sizeof(radix_AP_encodings[0]))));
+    }
 }
 
 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)
include/sysemu/kvm.h +1 -0
index 3045ee7678..01a8db1180 100644
@@ -526,5 +526,6 @@ int kvm_set_one_reg(CPUState *cs, uint64_t id, void *source);
  * Returns: 0 on success, or a negative errno on failure.
  */
 int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
+struct ppc_radix_page_info *kvm_get_radix_page_info(void);
 int kvm_get_max_memslots(void);
 #endif
target/ppc/cpu-qom.h +1 -0
index 4807f4d86c..0efb543912 100644
@@ -195,6 +195,7 @@ typedef struct PowerPCCPUClass {
     int bfd_mach;
     uint32_t l1_dcache_size, l1_icache_size;
     const struct ppc_segment_page_sizes *sps;
+    struct ppc_radix_page_info *radix_page_info;
     void (*init_proc)(CPUPPCState *env);
     int  (*check_pow)(CPUPPCState *env);
     int (*handle_mmu_fault)(PowerPCCPU *cpu, vaddr eaddr, int rwx, int mmu_idx);
target/ppc/cpu.h +4 -0
index b559b67073..a6c8c5ff4c 100644
@@ -934,6 +934,10 @@ struct ppc_segment_page_sizes {
     struct ppc_one_seg_page_size sps[PPC_PAGE_SIZES_MAX_SZ];
 };
 
+struct ppc_radix_page_info {
+    uint32_t count;
+    uint32_t entries[PPC_PAGE_SIZES_MAX_SZ];
+};
 
 /*****************************************************************************/
 /* The whole PowerPC CPU context */
target/ppc/kvm.c +25 -0
index d53ede8b4a..cf62a42c1f 100644
@@ -48,6 +48,7 @@
 #if defined(TARGET_PPC64)
 #include "hw/ppc/spapr_cpu_core.h"
 #endif
+#include "sysemu/kvm_int.h"
 
 //#define DEBUG_KVM
 
@@ -329,6 +330,30 @@ static void kvm_get_smmu_info(PowerPCCPU *cpu, struct kvm_ppc_smmu_info *info)
     kvm_get_fallback_smmu_info(cpu, info);
 }
 
+struct ppc_radix_page_info *kvm_get_radix_page_info(void)
+{
+    KVMState *s = KVM_STATE(current_machine->accelerator);
+    struct ppc_radix_page_info *radix_page_info;
+    struct kvm_ppc_rmmu_info rmmu_info;
+    int i;
+
+    if (!kvm_check_extension(s, KVM_CAP_PPC_MMU_RADIX)) {
+        return NULL;
+    }
+    if (kvm_vm_ioctl(s, KVM_PPC_GET_RMMU_INFO, &rmmu_info)) {
+        return NULL;
+    }
+    radix_page_info = g_malloc0(sizeof(*radix_page_info));
+    radix_page_info->count = 0;
+    for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
+        if (rmmu_info.ap_encodings[i]) {
+            radix_page_info->entries[i] = rmmu_info.ap_encodings[i];
+            radix_page_info->count++;
+        }
+    }
+    return radix_page_info;
+}
+
 static long gethugepagesize(const char *mem_path)
 {
     struct statfs fs;

target/ppc/kvm.c static void kvmppc_host_cpu_class_init(ObjectClass *oc, void *data) · line 2407

Quoted 4 lines
target/ppc/kvm.c +1 -0
@@ -2379,6 +2404,8 @@ static void kvmppc_host_cpu_class_init(ObjectClass *oc, void *data)
         pcc->l1_icache_size = icache_size;
     }
 
+    pcc->radix_page_info = kvm_enabled() ? kvm_get_radix_page_info() : NULL;

This whole function is only called in the kvm case: no need to check kvm_enabled() here.

Quoted 3 lines
target/ppc/kvm.c +0 -0
     /* Reason: kvmppc_host_cpu_initfn() dies when !kvm_enabled() */
     dc->cannot_destroy_with_object_finalize_yet = true;
 }
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
#13 lore raw

Re: [Qemu-devel] [RFC PATCH v2 06/12] spapr: Add ibm, processor-radix-AP-encodings to the device tree

Quoted 13 lines · David Gibson
Quoted 12 lines · Sam Bobroff

Use the new ioctl, KVM_PPC_GET_RMMU_INFO, to fetch radix MMU information from KVM and present the page encodings in the device tree
under ibm,processor-radix-AP-encodings. This provides page size information to the guest which is necessary for it to use radix mode.

---
v2:

Quoted 6 lines
Quoted 3 lines

* ppc_radix_page_info now kept in native format, conversion to BE done when adding to the device tree.
* radix_page_info moved into the CPU class, cleaning up some code.

Looks pretty good, although I imaginge it will need a little rework to
rebase on top of the TCG radix stuff.  Also one comment below..

I've reworked a bit so my TCG stuff applies cleanly on top.

Quoted 133 lines
Quoted 133 lines

 hw/ppc/spapr.c       | 12 ++++++++++++ include/sysemu/kvm.h |  1 + target/ppc/cpu-qom.h |  1 +
 target/ppc/cpu.h     |  4 ++++ target/ppc/kvm.c     | 27 +++++++++++++++++++++++++++

 5 files changed, 45 insertions(+)
hw/ppc/spapr.c +2 -0
index cceb35f083..ca3812555f 100644
@@ -409,6 +409,8 @@ static void spapr_populate_cpu_dt(CPUState *cs,
 void *fdt, int offset,
      sPAPRDRConnector *drc;
      sPAPRDRConnectorClass *drck;
      int drc_index;
+    uint32_t radix_AP_encodings[PPC_PAGE_SIZES_MAX_SZ];
+    int i;


     drc = spapr_dr_connector_by_id(SPAPR_DR_CONNECTOR_TYPE_CPU, index);
     if (drc) {

hw/ppc/spapr.c +0 -0
@@ -494,6 +496,16 @@ static void spapr_populate_cpu_dt(CPUState
 *cs, void *fdt, int offset,
      _FDT(spapr_fixup_cpu_numa_dt(fdt, offset, cs));


     _FDT(spapr_fixup_cpu_smt_dt(fdt, offset, cpu, compat_smt)); +
+    if (pcc->radix_page_info) {
+        for (i = 0; i < pcc->radix_page_info->count; i++) { +            radix_AP_encodings[i] = cpu_to_be32(pcc-

Quoted 1 line

radix_page_info->entries[i]);

+        }
+        _FDT((fdt_setprop(fdt, offset, "ibm,processor-radix-AP- encodings",
+                          radix_AP_encodings,
+                          pcc->radix_page_info->count * +                          sizeof(radix_AP_encodings[0])))); +    }
 }

 static void spapr_populate_cpus_dt_node(void *fdt, sPAPRMachineState *spapr)

include/sysemu/kvm.h +1 -0
index 3045ee7678..01a8db1180 100644
@@ -526,5 +526,6 @@ int kvm_set_one_reg(CPUState *cs, uint64_t id,
 void *source);
   * Returns: 0 on success, or a negative errno on failure.
   */
  int kvm_get_one_reg(CPUState *cs, uint64_t id, void *target);
+struct ppc_radix_page_info *kvm_get_radix_page_info(void);
  int kvm_get_max_memslots(void);

 #endif

target/ppc/cpu-qom.h +1 -0
index 4807f4d86c..0efb543912 100644
@@ -195,6 +195,7 @@ typedef struct PowerPCCPUClass {
      int bfd_mach;
      uint32_t l1_dcache_size, l1_icache_size;
      const struct ppc_segment_page_sizes *sps;
+    struct ppc_radix_page_info *radix_page_info;
      void (*init_proc)(CPUPPCState *env);
      int  (*check_pow)(CPUPPCState *env);
      int (*handle_mmu_fault)(PowerPCCPU *cpu, vaddr eaddr, int rwx,

int mmu_idx);

target/ppc/cpu.h +0 -0
index b559b67073..a6c8c5ff4c 100644
@@ -934,6 +934,10 @@ struct ppc_segment_page_sizes {
      struct ppc_one_seg_page_size sps[PPC_PAGE_SIZES_MAX_SZ];
  };

target/ppc/cpu.h +4 -0
+struct ppc_radix_page_info {
+    uint32_t count;
+    uint32_t entries[PPC_PAGE_SIZES_MAX_SZ];
+};

/***************************************************************** ************/
 /* The whole PowerPC CPU context */

target/ppc/kvm.c +1 -0
index d53ede8b4a..cf62a42c1f 100644
@@ -48,6 +48,7 @@
  #if defined(TARGET_PPC64)
  #include "hw/ppc/spapr_cpu_core.h"
  #endif
+#include "sysemu/kvm_int.h"


 //#define DEBUG_KVM

target/ppc/kvm.c +0 -0
@@ -329,6 +330,30 @@ static void kvm_get_smmu_info(PowerPCCPU *cpu,
 struct kvm_ppc_smmu_info *info)
      kvm_get_fallback_smmu_info(cpu, info);
  }

target/ppc/kvm.c +18 -0
+struct ppc_radix_page_info *kvm_get_radix_page_info(void)
+{
+    KVMState *s = KVM_STATE(current_machine->accelerator);
+    struct ppc_radix_page_info *radix_page_info;
+    struct kvm_ppc_rmmu_info rmmu_info;
+    int i;
+
+    if (!kvm_check_extension(s, KVM_CAP_PPC_MMU_RADIX)) {
+        return NULL;
+    }
+    if (kvm_vm_ioctl(s, KVM_PPC_GET_RMMU_INFO, &rmmu_info)) {
+        return NULL;
+    }
+    radix_page_info = g_malloc0(sizeof(*radix_page_info));
+    radix_page_info->count = 0;
+    for (i = 0; i < PPC_PAGE_SIZES_MAX_SZ; i++) {
+        if (rmmu_info.ap_encodings[i]) {
+            radix_page_info->entries[i] =

rmmu_info.ap_encodings[i];
+            radix_page_info->count++;
+        }
+    }
+    return radix_page_info;

target/ppc/kvm.c +2 -0
+}
+

 static long gethugepagesize(const char *mem_path) {
     struct statfs fs;

target/ppc/kvm.c +0 -0
@@ -2379,6 +2404,8 @@ static void
Quoted 8 lines
Quoted 6 lines
target/ppc/kvm.c +0 -0
 kvmppc_host_cpu_class_init(ObjectClass *oc, void *data)
          pcc->l1_icache_size = icache_size;
      }


+    pcc->radix_page_info = kvm_enabled() ?
kvm_get_radix_page_info() : NULL;

This whole function is only called in the kvm case: no need to check kvm_enabled() here.

I've reworked this, so this is irrelevant. I've added a generic function in mmu-radix64.h which calls kvm_get_radix_page_info if radix enabled, otherwise returns the default for TCG case.

Quoted 6 lines
Quoted 6 lines

     /* Reason: kvmppc_host_cpu_initfn() dies when !kvm_enabled() */
     dc->cannot_destroy_with_object_finalize_yet = true; }

#14 lore raw

RFC PATCH v2 7/12 target-ppc: support KVM_CAP_PPC_MMU_RADIX, KVM_CAP_PPC_MMU_HASH_V3

In reply to Sam Bobroff

Query and cache the value of two new KVM capabilities that indicate KVM's support for new radix and hash modes of the MMU.

---
v2:

* cap_mmu_hash renamed to cap_mmu_hash_v3.

 target/ppc/kvm.c     | 14 ++++++++++++++
 target/ppc/kvm_ppc.h | 12 ++++++++++++
 2 files changed, 26 insertions(+)
target/ppc/kvm.c +14 -0
index cf62a42c1f..8b153808fd 100644
@@ -83,6 +83,8 @@ static int cap_papr;
 static int cap_htab_fd;
 static int cap_fixup_hcalls;
 static int cap_htm;             /* Hardware transactional memory support */
+static int cap_mmu_radix;
+static int cap_mmu_hash_v3;
 
 static uint32_t debug_inst_opcode;
 
@@ -136,6 +138,8 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
     cap_htab_fd = kvm_check_extension(s, KVM_CAP_PPC_HTAB_FD);
     cap_fixup_hcalls = kvm_check_extension(s, KVM_CAP_PPC_FIXUP_HCALL);
     cap_htm = kvm_vm_check_extension(s, KVM_CAP_PPC_HTM);
+    cap_mmu_radix = kvm_vm_check_extension(s, KVM_CAP_PPC_MMU_RADIX);
+    cap_mmu_hash_v3 = kvm_vm_check_extension(s, KVM_CAP_PPC_MMU_HASH_V3);
 
     if (!cap_interrupt_level) {
         fprintf(stderr, "KVM: Couldn't find level irq capability. Expect the "
@@ -2430,6 +2434,16 @@ bool kvmppc_has_cap_htm(void)
     return cap_htm;
 }
 
+bool kvmppc_has_cap_mmu_radix(void)
+{
+    return cap_mmu_radix;
+}
+
+bool kvmppc_has_cap_mmu_hash_v3(void)
+{
+    return cap_mmu_hash_v3;
+}
+
 static PowerPCCPUClass *ppc_cpu_get_family_class(PowerPCCPUClass *pcc)
 {
     ObjectClass *oc = OBJECT_CLASS(pcc);
target/ppc/kvm_ppc.h +12 -0
index 8da2ee418a..56e222dfc2 100644
@@ -56,6 +56,8 @@ void kvmppc_hash64_write_pte(CPUPPCState *env, target_ulong pte_index,
                              target_ulong pte0, target_ulong pte1);
 bool kvmppc_has_cap_fixup_hcalls(void);
 bool kvmppc_has_cap_htm(void);
+bool kvmppc_has_cap_mmu_radix(void);
+bool kvmppc_has_cap_mmu_hash_v3(void);
 int kvmppc_enable_hwrng(void);
 int kvmppc_put_books_sregs(PowerPCCPU *cpu);
 PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
@@ -262,6 +264,16 @@ static inline bool kvmppc_has_cap_htm(void)
     return false;
 }
 
+static inline bool kvmppc_has_cap_mmu_radix(void)
+{
+    return false;
+}
+
+static inline bool kvmppc_has_cap_mmu_hash_v3(void)
+{
+    return false;
+}
+
 static inline int kvmppc_enable_hwrng(void)
 {
     return -1;
Signature · 1 line
2.11.0
Reviewed-by on v2 7/12
#15 lore raw

Re: [Qemu-devel] [RFC PATCH v2 07/12] target-ppc: support KVM_CAP_PPC_MMU_RADIX, KVM_CAP_PPC_MMU_HASH_V3

Quoted 4 lines · Sam Bobroff

Query and cache the value of two new KVM capabilities that indicate KVM's support for new radix and hash modes of the MMU.

Quoted 78 lines

---
v2:

* cap_mmu_hash renamed to cap_mmu_hash_v3.

 target/ppc/kvm.c     | 14 ++++++++++++++
 target/ppc/kvm_ppc.h | 12 ++++++++++++
 2 files changed, 26 insertions(+)
target/ppc/kvm.c +14 -0
index cf62a42c1f..8b153808fd 100644
@@ -83,6 +83,8 @@ static int cap_papr;
 static int cap_htab_fd;
 static int cap_fixup_hcalls;
 static int cap_htm;             /* Hardware transactional memory support */
+static int cap_mmu_radix;
+static int cap_mmu_hash_v3;
 
 static uint32_t debug_inst_opcode;
 
@@ -136,6 +138,8 @@ int kvm_arch_init(MachineState *ms, KVMState *s)
     cap_htab_fd = kvm_check_extension(s, KVM_CAP_PPC_HTAB_FD);
     cap_fixup_hcalls = kvm_check_extension(s, KVM_CAP_PPC_FIXUP_HCALL);
     cap_htm = kvm_vm_check_extension(s, KVM_CAP_PPC_HTM);
+    cap_mmu_radix = kvm_vm_check_extension(s, KVM_CAP_PPC_MMU_RADIX);
+    cap_mmu_hash_v3 = kvm_vm_check_extension(s, KVM_CAP_PPC_MMU_HASH_V3);
 
     if (!cap_interrupt_level) {
         fprintf(stderr, "KVM: Couldn't find level irq capability. Expect the "
@@ -2430,6 +2434,16 @@ bool kvmppc_has_cap_htm(void)
     return cap_htm;
 }
 
+bool kvmppc_has_cap_mmu_radix(void)
+{
+    return cap_mmu_radix;
+}
+
+bool kvmppc_has_cap_mmu_hash_v3(void)
+{
+    return cap_mmu_hash_v3;
+}
+
 static PowerPCCPUClass *ppc_cpu_get_family_class(PowerPCCPUClass *pcc)
 {
     ObjectClass *oc = OBJECT_CLASS(pcc);
target/ppc/kvm_ppc.h +12 -0
index 8da2ee418a..56e222dfc2 100644
@@ -56,6 +56,8 @@ void kvmppc_hash64_write_pte(CPUPPCState *env, target_ulong pte_index,
                              target_ulong pte0, target_ulong pte1);
 bool kvmppc_has_cap_fixup_hcalls(void);
 bool kvmppc_has_cap_htm(void);
+bool kvmppc_has_cap_mmu_radix(void);
+bool kvmppc_has_cap_mmu_hash_v3(void);
 int kvmppc_enable_hwrng(void);
 int kvmppc_put_books_sregs(PowerPCCPU *cpu);
 PowerPCCPUClass *kvm_ppc_get_host_cpu_class(void);
@@ -262,6 +264,16 @@ static inline bool kvmppc_has_cap_htm(void)
     return false;
 }
 
+static inline bool kvmppc_has_cap_mmu_radix(void)
+{
+    return false;
+}
+
+static inline bool kvmppc_has_cap_mmu_hash_v3(void)
+{
+    return false;
+}
+
 static inline int kvmppc_enable_hwrng(void)
 {
     return -1;
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
#16 lore raw

RFC PATCH v2 8/12 spapr: Only setup HTP if necessary.

In reply to Sam Bobroff

If QEMU is using KVM, and KVM is capable of running in radix mode, guests can be run in real-mode without allocating a HPT (because KVM will use a minimal RPT). So in this case, we avoid creating the HPT at reset time and later (during CAS) create it if it is necessary.

---
v2:

* This patch has been mostly rewritten to move the late HPT allocation to CAS. This allows a guest to start in radix mode (when it's in real mode) and then change to hash, even if it is a legacy guest and will not call h_register_process_table().
* Added an exported function to spapr.c to perform HPT allocation and adjust the vrma if necessary. This makes it possible to allocate the HPT from h_client_architecture_support() in spapr_hcall.c.

 hw/ppc/spapr.c         | 24 +++++++++++++++---------
 hw/ppc/spapr_hcall.c   | 10 ++++++++++
 include/hw/ppc/spapr.h |  1 +
 3 files changed, 26 insertions(+), 9 deletions(-)
hw/ppc/spapr.c +15 -9
index ca3812555f..dfee0f685f 100644
@@ -1123,6 +1123,17 @@ static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
     }
 }
 
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr)
+{
+    spapr_reallocate_hpt(spapr,
+                     spapr_hpt_shift_for_ramsize(MACHINE(qdev_get_machine())->maxram_size),
+                     &error_fatal);
+    if (spapr->vrma_adjust) {
+        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
+                                          spapr->htab_shift);
+    }
+}
+
 static void find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
 {
     bool matched = false;
@@ -1151,15 +1162,10 @@ static void ppc_spapr_reset(void)
     /* Check for unknown sysbus devices */
     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
 
-    /* Allocate and/or reset the hash page table */
-    spapr_reallocate_hpt(spapr,
-                         spapr_hpt_shift_for_ramsize(machine->maxram_size),
-                         &error_fatal);
-
-    /* Update the RMA size if necessary */
-    if (spapr->vrma_adjust) {
-        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
-                                          spapr->htab_shift);
+    /* If using KVM with radix mode available, VCPUs can be started
+     * without a HPT because KVM will start them in radix mode. */
+    if (!(kvm_enabled() && kvmppc_has_cap_mmu_radix())) {
+        spapr_setup_hpt_and_vrma(spapr);
     }
 
     qemu_devices_reset();
hw/ppc/spapr_hcall.c +10 -0
index 42d20e0b92..cea34073aa 100644
@@ -1002,6 +1002,16 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     ov5_updates = spapr_ovec_new();
     spapr->cas_reboot = spapr_ovec_diff(ov5_updates,
                                         ov5_cas_old, spapr->ov5_cas);
+    if (kvm_enabled()) {
+        if (kvmppc_has_cap_mmu_radix()) {
+            /* If the HPT hasn't yet been set up (see
+             * ppc_spapr_reset()), and it's needed, do it now: */
+            if (!spapr_ovec_test(ov5_updates, OV5_MMU_RADIX)) {
+                /* legacy hash or new hash: */
+                spapr_setup_hpt_and_vrma(spapr);
+            }
+        }
+    }
 
     if (!spapr->cas_reboot) {
         spapr->cas_reboot =
include/hw/ppc/spapr.h +1 -0
index f9b17d860a..a30cbc485c 100644
@@ -590,6 +590,7 @@ void spapr_dt_events(sPAPRMachineState *sm, void *fdt);
 int spapr_h_cas_compose_response(sPAPRMachineState *sm,
                                  target_ulong addr, target_ulong size,
                                  sPAPROptionVector *ov5_updates);
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr);
 sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
 void spapr_tce_table_enable(sPAPRTCETable *tcet,
                             uint32_t page_shift, uint64_t bus_offset,
Signature · 1 line
2.11.0
#17 lore raw

Re: [Qemu-devel] [RFC PATCH v2 08/12] spapr: Only setup HTP if necessary.

s/HTP/HPT/ in subject line.

Quoted 6 lines · Sam Bobroff

If QEMU is using KVM, and KVM is capable of running in radix mode, guests can be run in real-mode without allocating a HPT (because KVM will use a minimal RPT). So in this case, we avoid creating the HPT at reset time and later (during CAS) create it if it is necessary.

So, IIRC, we discussed previously that the logical way to do things was to, by default, delay HPT allocation until CAS time, and just do it at reset time for the case that needs it: hash host with KVM.

Did you hit a problem with that approach, or is there still work to be done here?

Quoted 63 lines

---
v2:

* This patch has been mostly rewritten to move the late HPT allocation to CAS. This allows a guest to start in radix mode (when it's in real mode) and then change to hash, even if it is a legacy guest and will not call h_register_process_table().
* Added an exported function to spapr.c to perform HPT allocation and adjust the vrma if necessary. This makes it possible to allocate the HPT from h_client_architecture_support() in spapr_hcall.c.

 hw/ppc/spapr.c         | 24 +++++++++++++++---------
 hw/ppc/spapr_hcall.c   | 10 ++++++++++
 include/hw/ppc/spapr.h |  1 +
 3 files changed, 26 insertions(+), 9 deletions(-)
hw/ppc/spapr.c +15 -9
index ca3812555f..dfee0f685f 100644
@@ -1123,6 +1123,17 @@ static void spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
     }
 }
 
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr)
+{
+    spapr_reallocate_hpt(spapr,
+                     spapr_hpt_shift_for_ramsize(MACHINE(qdev_get_machine())->maxram_size),
+                     &error_fatal);
+    if (spapr->vrma_adjust) {
+        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
+                                          spapr->htab_shift);
+    }
+}
+
 static void find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
 {
     bool matched = false;
@@ -1151,15 +1162,10 @@ static void ppc_spapr_reset(void)
     /* Check for unknown sysbus devices */
     foreach_dynamic_sysbus_device(find_unknown_sysbus_device, NULL);
 
-    /* Allocate and/or reset the hash page table */
-    spapr_reallocate_hpt(spapr,
-                         spapr_hpt_shift_for_ramsize(machine->maxram_size),
-                         &error_fatal);
-
-    /* Update the RMA size if necessary */
-    if (spapr->vrma_adjust) {
-        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(),
-                                          spapr->htab_shift);
+    /* If using KVM with radix mode available, VCPUs can be started
+     * without a HPT because KVM will start them in radix mode. */
+    if (!(kvm_enabled() && kvmppc_has_cap_mmu_radix())) {
+        spapr_setup_hpt_and_vrma(spapr);
     }
 
     qemu_devices_reset();
hw/ppc/spapr_hcall.c +0 -0
index 42d20e0b92..cea34073aa 100644

hw/ppc/spapr_hcall.c static target_ulong h_client_architecture_support(PowerPCCPU *cpu, · line 1008

Quoted 7 lines
hw/ppc/spapr_hcall.c +4 -0
@@ -1002,6 +1002,16 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     ov5_updates = spapr_ovec_new();
     spapr->cas_reboot = spapr_ovec_diff(ov5_updates,
                                         ov5_cas_old, spapr->ov5_cas);
+    if (kvm_enabled()) {
+        if (kvmppc_has_cap_mmu_radix()) {
+            /* If the HPT hasn't yet been set up (see
+             * ppc_spapr_reset()), and it's needed, do it now: */

I think it's a bit fragile to have here it explicitly mirror the logic which determines whether the HPT is allocated early. I'd prefer to explicitly test here whether we have allocated an HPT - adding a flag, if we have to.

Quoted 21 lines
hw/ppc/spapr_hcall.c +6 -0
+            if (!spapr_ovec_test(ov5_updates, OV5_MMU_RADIX)) {
+                /* legacy hash or new hash: */
+                spapr_setup_hpt_and_vrma(spapr);
+            }
+        }
+    }
 
     if (!spapr->cas_reboot) {
         spapr->cas_reboot =
include/hw/ppc/spapr.h +1 -0
index f9b17d860a..a30cbc485c 100644
@@ -590,6 +590,7 @@ void spapr_dt_events(sPAPRMachineState *sm, void *fdt);
 int spapr_h_cas_compose_response(sPAPRMachineState *sm,
                                  target_ulong addr, target_ulong size,
                                  sPAPROptionVector *ov5_updates);
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr);
 sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
 void spapr_tce_table_enable(sPAPRTCETable *tcet,
                             uint32_t page_shift, uint64_t bus_offset,
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
#18 lore raw

Re: [Qemu-devel] [RFC PATCH v2 08/12] spapr: Only setup HTP if necessary.

Quoted 11 lines · David Gibson

s/HTP/HPT/ in subject line.

Quoted 7 lines · Sam Bobroff

If QEMU is using KVM, and KVM is capable of running in radix mode, guests can be run in real-mode without allocating a HPT (because KVM
will use a minimal RPT). So in this case, we avoid creating the HPT at reset time and later (during CAS) create it if it is necessary.

Quoted 8 lines
Quoted 1 line

So, IIRC, we discussed previously that the logical way to do things was to, by default, delay HPT allocation until CAS time, and just do it at reset time for the case that needs it: hash host with KVM.

Did you hit a problem with that approach, or is there still work to be
done here?

So what we're doing is assuming radix. Allocate hpt if hash host, otherwise delay til CAS time and allocate only if guest chose hash.

Quoted 84 lines
Quoted 79 lines

---
v2:

* This patch has been mostly rewritten to move the late HPT allocation to CAS.
This allows a guest to start in radix mode (when it's in real mode) and then
change to hash, even if it is a legacy guest and will not call h_register_process_table().
* Added an exported function to spapr.c to perform HPT allocation and adjust
the vrma if necessary. This makes it possible to allocate the HPT from
h_client_architecture_support() in spapr_hcall.c.

 hw/ppc/spapr.c         | 24 +++++++++++++++--------- hw/ppc/spapr_hcall.c   | 10 ++++++++++ include/hw/ppc/spapr.h |  1 +

 3 files changed, 26 insertions(+), 9 deletions(-)
hw/ppc/spapr.c +0 -0
index ca3812555f..dfee0f685f 100644
@@ -1123,6 +1123,17 @@ static void
 spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
      }
  }

hw/ppc/spapr.c +4 -0
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr)
+{
+    spapr_reallocate_hpt(spapr,
+                     spapr_hpt_shift_for_ramsize(MACHINE(qdev_get_

machine())->maxram_size),
+                     &error_fatal);
+    if (spapr->vrma_adjust) {
+        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), +                                          spapr->htab_shift); +    }

hw/ppc/spapr.c +2 -0
+}
+

 static void find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
 {
     bool matched = false;

hw/ppc/spapr.c +0 -0
@@ -1151,15 +1162,10 @@ static void ppc_spapr_reset(void)
      /* Check for unknown sysbus devices */
      foreach_dynamic_sysbus_device(find_unknown_sysbus_device,
 NULL);


-    /* Allocate and/or reset the hash page table */ -    spapr_reallocate_hpt(spapr,
-                         spapr_hpt_shift_for_ramsize(machine-

Quoted 1 line

maxram_size),

-                         &error_fatal);
-
-    /* Update the RMA size if necessary */
-    if (spapr->vrma_adjust) {
-        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), -                                          spapr->htab_shift); +    /* If using KVM with radix mode available, VCPUs can be started
+     * without a HPT because KVM will start them in radix mode. */ +    if (!(kvm_enabled() && kvmppc_has_cap_mmu_radix())) { +        spapr_setup_hpt_and_vrma(spapr);
     }

     qemu_devices_reset();

hw/ppc/spapr_hcall.c +0 -0
index 42d20e0b92..cea34073aa 100644
@@ -1002,6 +1002,16 @@ static target_ulong
 h_client_architecture_support(PowerPCCPU *cpu,
      ov5_updates = spapr_ovec_new();
      spapr->cas_reboot = spapr_ovec_diff(ov5_updates,

                                         ov5_cas_old, spapr-

Quoted 5 lines
Quoted 1 line

ov5_cas);

+    if (kvm_enabled()) {
+        if (kvmppc_has_cap_mmu_radix()) {
+            /* If the HPT hasn't yet been set up (see +             * ppc_spapr_reset()), and it's needed, do it now: */

Quoted 6 lines

I think it's a bit fragile to have here it explicitly mirror the logic
which determines whether the HPT is allocated early.  I'd prefer to explicitly test here whether we have allocated an HPT - adding a flag,
if we have to.

We can use the MSB of patb_entry as that flag.

patb_entry & GUEST_RADIX == GUEST_RADIX -> radix, so assume a hpt hasn't been allocated.

When we do allocate a hpt we know we're not radix, so set patb_entry &= ~GUEST_RADIX;

Where GUEST_RADIX is the msb in patb_entry which indicates that a guest is radix.

Essentially patb_entry & GUEST_RADIX cleared mean hash with hpt allocated, patb_entry & GUEST_RADIX set means radix so assume an hpt hasn't been allocated. On the hpt allocation path we clear GUEST_RADIX in patb_entry and when we set GUEST_RADIX we free the hpt.

Quoted 27 lines
Quoted 27 lines

+            if (!spapr_ovec_test(ov5_updates, OV5_MMU_RADIX)) { +                /* legacy hash or new hash: */
+                spapr_setup_hpt_and_vrma(spapr);
+            }
+        }
+    }

     if (!spapr->cas_reboot) {
         spapr->cas_reboot =

include/hw/ppc/spapr.h +0 -0
index f9b17d860a..a30cbc485c 100644
@@ -590,6 +590,7 @@ void spapr_dt_events(sPAPRMachineState *sm,
 void *fdt);
  int spapr_h_cas_compose_response(sPAPRMachineState *sm,

                                  target_ulong addr, target_ulong size,
                                  sPAPROptionVector *ov5_updates); +void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr); sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
 void spapr_tce_table_enable(sPAPRTCETable *tcet, uint32_t page_shift, uint64_t bus_offset,

#19 lore raw

Re: [Qemu-devel] [RFC PATCH v2 08/12] spapr: Only setup HTP if necessary.

Quoted 108 lines · Suraj Jitindar Singh
Quoted 11 lines · David Gibson

s/HTP/HPT/ in subject line.

Quoted 7 lines · Sam Bobroff

If QEMU is using KVM, and KVM is capable of running in radix mode, guests can be run in real-mode without allocating a HPT (because KVM
will use a minimal RPT). So in this case, we avoid creating the HPT at reset time and later (during CAS) create it if it is necessary.

Quoted 8 lines
Quoted 1 line

So, IIRC, we discussed previously that the logical way to do things was to, by default, delay HPT allocation until CAS time, and just do it at reset time for the case that needs it: hash host with KVM.

Did you hit a problem with that approach, or is there still work to be
done here?

So what we're doing is assuming radix. Allocate hpt if hash host, otherwise delay til CAS time and allocate only if guest chose hash.

Quoted 84 lines
Quoted 84 lines

---
v2:

* This patch has been mostly rewritten to move the late HPT allocation to CAS.
This allows a guest to start in radix mode (when it's in real mode) and then
change to hash, even if it is a legacy guest and will not call h_register_process_table().
* Added an exported function to spapr.c to perform HPT allocation and adjust
the vrma if necessary. This makes it possible to allocate the HPT from
h_client_architecture_support() in spapr_hcall.c.

 hw/ppc/spapr.c         | 24 +++++++++++++++--------- hw/ppc/spapr_hcall.c   | 10 ++++++++++ include/hw/ppc/spapr.h |  1 +

 3 files changed, 26 insertions(+), 9 deletions(-)
hw/ppc/spapr.c +0 -0
index ca3812555f..dfee0f685f 100644
@@ -1123,6 +1123,17 @@ static void
 spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
      }
  }

hw/ppc/spapr.c +4 -0
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr)
+{
+    spapr_reallocate_hpt(spapr,
+                     spapr_hpt_shift_for_ramsize(MACHINE(qdev_get_

machine())->maxram_size),
+                     &error_fatal);
+    if (spapr->vrma_adjust) {
+        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), +                                          spapr->htab_shift); +    }

hw/ppc/spapr.c +2 -0
+}
+

 static void find_unknown_sysbus_device(SysBusDevice *sbdev, void *opaque)
 {
     bool matched = false;

hw/ppc/spapr.c +0 -0
@@ -1151,15 +1162,10 @@ static void ppc_spapr_reset(void)
      /* Check for unknown sysbus devices */
      foreach_dynamic_sysbus_device(find_unknown_sysbus_device,
 NULL);


-    /* Allocate and/or reset the hash page table */ -    spapr_reallocate_hpt(spapr,
-                         spapr_hpt_shift_for_ramsize(machine-

Quoted 1 line

maxram_size),

-                         &error_fatal);
-
-    /* Update the RMA size if necessary */
-    if (spapr->vrma_adjust) {
-        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), -                                          spapr->htab_shift); +    /* If using KVM with radix mode available, VCPUs can be started
+     * without a HPT because KVM will start them in radix mode. */ +    if (!(kvm_enabled() && kvmppc_has_cap_mmu_radix())) { +        spapr_setup_hpt_and_vrma(spapr);
     }

     qemu_devices_reset();

hw/ppc/spapr_hcall.c +0 -0
index 42d20e0b92..cea34073aa 100644
@@ -1002,6 +1002,16 @@ static target_ulong
 h_client_architecture_support(PowerPCCPU *cpu,
      ov5_updates = spapr_ovec_new();
      spapr->cas_reboot = spapr_ovec_diff(ov5_updates,

                                         ov5_cas_old, spapr-

Quoted 1 line

ov5_cas);

+    if (kvm_enabled()) {
+        if (kvmppc_has_cap_mmu_radix()) {
+            /* If the HPT hasn't yet been set up (see +             * ppc_spapr_reset()), and it's needed, do it now: */

Quoted 8 lines
Quoted 6 lines

I think it's a bit fragile to have here it explicitly mirror the logic
which determines whether the HPT is allocated early.  I'd prefer to explicitly test here whether we have allocated an HPT - adding a flag,
if we have to.

We can use the MSB of patb_entry as that flag.

Uh.. only for POWER9..

Quoted 42 lines

patb_entry & GUEST_RADIX == GUEST_RADIX -> radix, so assume a hpt hasn't been allocated.

When we do allocate a hpt we know we're not radix, so set patb_entry &= ~GUEST_RADIX;

Where GUEST_RADIX is the msb in patb_entry which indicates that a guest is radix.

Essentially patb_entry & GUEST_RADIX cleared mean hash with hpt allocated, patb_entry & GUEST_RADIX set means radix so assume an hpt hasn't been allocated. On the hpt allocation path we clear GUEST_RADIX in patb_entry and when we set GUEST_RADIX we free the hpt.

Quoted 27 lines
Quoted 27 lines

+            if (!spapr_ovec_test(ov5_updates, OV5_MMU_RADIX)) { +                /* legacy hash or new hash: */
+                spapr_setup_hpt_and_vrma(spapr);
+            }
+        }
+    }

     if (!spapr->cas_reboot) {
         spapr->cas_reboot =

include/hw/ppc/spapr.h +0 -0
index f9b17d860a..a30cbc485c 100644
@@ -590,6 +590,7 @@ void spapr_dt_events(sPAPRMachineState *sm,
 void *fdt);
  int spapr_h_cas_compose_response(sPAPRMachineState *sm,

                                  target_ulong addr, target_ulong size,
                                  sPAPROptionVector *ov5_updates); +void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr); sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t liobn);
 void spapr_tce_table_enable(sPAPRTCETable *tcet, uint32_t page_shift, uint64_t bus_offset,

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
#20 lore raw

Re: [Qemu-devel] [RFC PATCH v2 08/12] spapr: Only setup HTP if necessary.

Quoted 136 lines · David Gibson
Quoted 135 lines · Suraj Jitindar Singh
Quoted 24 lines · David Gibson

s/HTP/HPT/ in subject line.

Quoted 12 lines · Sam Bobroff

If QEMU is using KVM, and KVM is capable of running in radix mode,
guests can be run in real-mode without allocating a HPT (because
KVM
will use a minimal RPT). So in this case, we avoid creating the HPT
at reset time and later (during CAS) create it if it is necessary.

Quoted 1 line

So, IIRC, we discussed previously that the logical way to do things
was to, by default, delay HPT allocation until CAS time, and just do
it at reset time for the case that needs it: hash host with KVM.

Quoted 4 lines

Did you hit a problem with that approach, or is there still work to
be
done here?

So what we're doing is assuming radix. Allocate hpt if hash host, otherwise delay til CAS time and allocate only if guest chose hash.

Quoted 102 lines
Quoted 89 lines

---
v2:

* This patch has been mostly rewritten to move the late HPT allocation to CAS.
This allows a guest to start in radix mode (when it's in real mode)
and then
change to hash, even if it is a legacy guest and will not call h_register_process_table().
* Added an exported function to spapr.c to perform HPT allocation
and adjust
the vrma if necessary. This makes it possible to allocate the HPT
from
h_client_architecture_support() in spapr_hcall.c.

 hw/ppc/spapr.c         | 24 +++++++++++++++--------- hw/ppc/spapr_hcall.c   | 10 ++++++++++ include/hw/ppc/spapr.h |  1 +

 3 files changed, 26 insertions(+), 9 deletions(-)
hw/ppc/spapr.c +0 -0
index ca3812555f..dfee0f685f 100644
@@ -1123,6 +1123,17 @@ static void
 spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
      }
  }

hw/ppc/spapr.c +4 -0
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr)
+{
+    spapr_reallocate_hpt(spapr,
+                     spapr_hpt_shift_for_ramsize(MACHINE(qdev_

get_
machine())->maxram_size),
+                     &error_fatal);
+    if (spapr->vrma_adjust) {
+        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), +                                          spapr->htab_shift); +    }

hw/ppc/spapr.c +2 -0
+}
+

 static void find_unknown_sysbus_device(SysBusDevice *sbdev, void
*opaque)
 {
     bool matched = false;

hw/ppc/spapr.c +0 -0
@@ -1151,15 +1162,10 @@ static void ppc_spapr_reset(void)
      /* Check for unknown sysbus devices */
      foreach_dynamic_sysbus_device(find_unknown_sysbus_device,
 NULL);


-    /* Allocate and/or reset the hash page table */ -    spapr_reallocate_hpt(spapr,
-                         spapr_hpt_shift_for_ramsize(machine-

Quoted 2 lines

maxram_size),

-                         &error_fatal);
-
-    /* Update the RMA size if necessary */
-    if (spapr->vrma_adjust) {
-        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), -                                          spapr->htab_shift); +    /* If using KVM with radix mode available, VCPUs can be started
+     * without a HPT because KVM will start them in radix mode. */
+    if (!(kvm_enabled() && kvmppc_has_cap_mmu_radix())) { +        spapr_setup_hpt_and_vrma(spapr);
     }

     qemu_devices_reset();

hw/ppc/spapr_hcall.c +0 -0
index 42d20e0b92..cea34073aa 100644
@@ -1002,6 +1002,16 @@ static target_ulong
 h_client_architecture_support(PowerPCCPU *cpu,
      ov5_updates = spapr_ovec_new();
      spapr->cas_reboot = spapr_ovec_diff(ov5_updates,

                                         ov5_cas_old, spapr-

Quoted 1 line
Quoted 6 lines
Quoted 1 line

ov5_cas);

+    if (kvm_enabled()) {
+        if (kvmppc_has_cap_mmu_radix()) {
+            /* If the HPT hasn't yet been set up (see +             * ppc_spapr_reset()), and it's needed, do it now:
*/

I think it's a bit fragile to have here it explicitly mirror the logic
which determines whether the HPT is allocated early.  I'd prefer to
explicitly test here whether we have allocated an HPT - adding a flag,
if we have to.

Quoted 2 lines
Quoted 1 line

We can use the MSB of patb_entry as that flag.

Uh.. only for POWER9..

Well on <POWER9, patb_entry will always be zero, which we're taking to mean a hpt has been allocated.

We could always just check spapr->htab == NULL...

Quoted 52 lines
Quoted 52 lines

patb_entry & GUEST_RADIX == GUEST_RADIX -> radix, so assume a hpt hasn't been allocated.

When we do allocate a hpt we know we're not radix, so set patb_entry &= ~GUEST_RADIX;

Where GUEST_RADIX is the msb in patb_entry which indicates that a guest
is radix.

Essentially patb_entry & GUEST_RADIX cleared mean hash with hpt allocated, patb_entry & GUEST_RADIX set means radix so assume an hpt
hasn't been allocated. On the hpt allocation path we clear GUEST_RADIX
in patb_entry and when we set GUEST_RADIX we free the hpt.

Quoted 33 lines
Quoted 33 lines

+            if (!spapr_ovec_test(ov5_updates, OV5_MMU_RADIX)) {
+                /* legacy hash or new hash: */
+                spapr_setup_hpt_and_vrma(spapr);
+            }
+        }
+    }

     if (!spapr->cas_reboot) {
         spapr->cas_reboot =

include/hw/ppc/spapr.h +0 -0
index f9b17d860a..a30cbc485c 100644
@@ -590,6 +590,7 @@ void spapr_dt_events(sPAPRMachineState *sm,
 void *fdt);
  int spapr_h_cas_compose_response(sPAPRMachineState *sm,
                                   target_ulong addr,

target_ulong
size,
                                  sPAPROptionVector *ov5_updates);
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr); sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t
liobn);
 void spapr_tce_table_enable(sPAPRTCETable *tcet, uint32_t page_shift, uint64_t bus_offset,

#21 lore raw

Re: [Qemu-devel] [RFC PATCH v2 08/12] spapr: Only setup HTP if necessary.

Quoted 137 lines · Suraj Jitindar Singh
Quoted 136 lines · David Gibson
Quoted 135 lines · Suraj Jitindar Singh
Quoted 24 lines · David Gibson

s/HTP/HPT/ in subject line.

Quoted 12 lines · Sam Bobroff

If QEMU is using KVM, and KVM is capable of running in radix mode,
guests can be run in real-mode without allocating a HPT (because
KVM
will use a minimal RPT). So in this case, we avoid creating the HPT
at reset time and later (during CAS) create it if it is necessary.

Quoted 1 line

So, IIRC, we discussed previously that the logical way to do things
was to, by default, delay HPT allocation until CAS time, and just do
it at reset time for the case that needs it: hash host with KVM.

Quoted 4 lines

Did you hit a problem with that approach, or is there still work to
be
done here?

So what we're doing is assuming radix. Allocate hpt if hash host, otherwise delay til CAS time and allocate only if guest chose hash.

Quoted 102 lines
Quoted 89 lines

---
v2:

* This patch has been mostly rewritten to move the late HPT allocation to CAS.
This allows a guest to start in radix mode (when it's in real mode)
and then
change to hash, even if it is a legacy guest and will not call h_register_process_table().
* Added an exported function to spapr.c to perform HPT allocation
and adjust
the vrma if necessary. This makes it possible to allocate the HPT
from
h_client_architecture_support() in spapr_hcall.c.

 hw/ppc/spapr.c         | 24 +++++++++++++++--------- hw/ppc/spapr_hcall.c   | 10 ++++++++++ include/hw/ppc/spapr.h |  1 +

 3 files changed, 26 insertions(+), 9 deletions(-)
hw/ppc/spapr.c +0 -0
index ca3812555f..dfee0f685f 100644
@@ -1123,6 +1123,17 @@ static void
 spapr_reallocate_hpt(sPAPRMachineState *spapr, int shift,
      }
  }

hw/ppc/spapr.c +4 -0
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr)
+{
+    spapr_reallocate_hpt(spapr,
+                     spapr_hpt_shift_for_ramsize(MACHINE(qdev_

get_
machine())->maxram_size),
+                     &error_fatal);
+    if (spapr->vrma_adjust) {
+        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), +                                          spapr->htab_shift); +    }

hw/ppc/spapr.c +2 -0
+}
+

 static void find_unknown_sysbus_device(SysBusDevice *sbdev, void
*opaque)
 {
     bool matched = false;

hw/ppc/spapr.c +0 -0
@@ -1151,15 +1162,10 @@ static void ppc_spapr_reset(void)
      /* Check for unknown sysbus devices */
      foreach_dynamic_sysbus_device(find_unknown_sysbus_device,
 NULL);


-    /* Allocate and/or reset the hash page table */ -    spapr_reallocate_hpt(spapr,
-                         spapr_hpt_shift_for_ramsize(machine-

Quoted 2 lines

maxram_size),

-                         &error_fatal);
-
-    /* Update the RMA size if necessary */
-    if (spapr->vrma_adjust) {
-        spapr->rma_size = kvmppc_rma_size(spapr_node0_size(), -                                          spapr->htab_shift); +    /* If using KVM with radix mode available, VCPUs can be started
+     * without a HPT because KVM will start them in radix mode. */
+    if (!(kvm_enabled() && kvmppc_has_cap_mmu_radix())) { +        spapr_setup_hpt_and_vrma(spapr);
     }

     qemu_devices_reset();

hw/ppc/spapr_hcall.c +0 -0
index 42d20e0b92..cea34073aa 100644
@@ -1002,6 +1002,16 @@ static target_ulong
 h_client_architecture_support(PowerPCCPU *cpu,
      ov5_updates = spapr_ovec_new();
      spapr->cas_reboot = spapr_ovec_diff(ov5_updates,

                                         ov5_cas_old, spapr-

Quoted 1 line
Quoted 6 lines
Quoted 1 line

ov5_cas);

+    if (kvm_enabled()) {
+        if (kvmppc_has_cap_mmu_radix()) {
+            /* If the HPT hasn't yet been set up (see +             * ppc_spapr_reset()), and it's needed, do it now:
*/

I think it's a bit fragile to have here it explicitly mirror the logic
which determines whether the HPT is allocated early.  I'd prefer to
explicitly test here whether we have allocated an HPT - adding a flag,
if we have to.

Quoted 5 lines
Quoted 2 lines
Quoted 1 line

We can use the MSB of patb_entry as that flag.

Uh.. only for POWER9..

Well on <POWER9, patb_entry will always be zero, which we're taking to mean a hpt has been allocated.

Hrm, I suppose so.

Quoted 1 line

We could always just check spapr->htab == NULL...

Ah.. no. Because it will be NULL when there is an HPT, but it's inside KVM.

Quoted 54 lines
Quoted 53 lines
Quoted 53 lines

patb_entry & GUEST_RADIX == GUEST_RADIX -> radix, so assume a hpt hasn't been allocated.

When we do allocate a hpt we know we're not radix, so set patb_entry &= ~GUEST_RADIX;

Where GUEST_RADIX is the msb in patb_entry which indicates that a guest
is radix.

Essentially patb_entry & GUEST_RADIX cleared mean hash with hpt allocated, patb_entry & GUEST_RADIX set means radix so assume an hpt
hasn't been allocated. On the hpt allocation path we clear GUEST_RADIX
in patb_entry and when we set GUEST_RADIX we free the hpt.

Quoted 33 lines
Quoted 33 lines

+            if (!spapr_ovec_test(ov5_updates, OV5_MMU_RADIX)) {
+                /* legacy hash or new hash: */
+                spapr_setup_hpt_and_vrma(spapr);
+            }
+        }
+    }

     if (!spapr->cas_reboot) {
         spapr->cas_reboot =

include/hw/ppc/spapr.h +0 -0
index f9b17d860a..a30cbc485c 100644
@@ -590,6 +590,7 @@ void spapr_dt_events(sPAPRMachineState *sm,
 void *fdt);
  int spapr_h_cas_compose_response(sPAPRMachineState *sm,
                                   target_ulong addr,

target_ulong
size,
                                  sPAPROptionVector *ov5_updates);
+void spapr_setup_hpt_and_vrma(sPAPRMachineState *spapr); sPAPRTCETable *spapr_tce_new_table(DeviceState *owner, uint32_t
liobn);
 void spapr_tce_table_enable(sPAPRTCETable *tcet, uint32_t page_shift, uint64_t bus_offset,

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
#22 lore raw

RFC PATCH v2 9/12 spapr: Add h_register_process_table() hypercall

In reply to Sam Bobroff

Both radix and hash modes require guests to use h_register_process_table() to set up the MMU. Implement it using the new KVM ioctl KVM_PPC_CONFIGURE_V3_MMU.

This hypercall is also necessary for fully emulated guests, so it will need to be reworked to integrate with Suraj's TCG patchset.
---
v2:

* I haven't addressed review comments for this patch because it overlaps with Suraj's implementation of the same function and we'll work together to integrate them.

 hw/ppc/spapr_hcall.c   | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
 include/hw/ppc/spapr.h |  1 +
 target/ppc/kvm.c       | 12 ++++++++++++
 target/ppc/kvm_ppc.h   |  1 +
 4 files changed, 62 insertions(+)
hw/ppc/spapr_hcall.c +48 -0
index cea34073aa..9391619ed6 100644
@@ -1027,6 +1027,50 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     return H_SUCCESS;
 }
 
+static target_ulong h_register_process_table(PowerPCCPU *cpu,
+                                             sPAPRMachineState *spapr,
+                                             target_ulong opcode,
+                                             target_ulong *args)
+{
+    static target_ulong last_process_table;
+    target_ulong flags = args[0];
+    target_ulong proc_tbl = args[1];
+    target_ulong page_size = args[2];
+    target_ulong table_size = args[3];
+    uint64_t cflags, cproc;
+
+    cflags = (flags & 4) ? KVM_PPC_MMUV3_RADIX : 0;
+    cflags |= (flags & 1) ? KVM_PPC_MMUV3_GTSE : 0;
+    cproc = (flags & 4) ? (1ul << 63) : 0;
+    if (!(flags & 0x10)) {
+        if ((last_process_table & (1ul << 63)) != cproc) {
+            return H_PARAMETER;
+        }
+        cproc = last_process_table;
+    } else if (!(flags & 0x8)) {
+        ; /* do nothing */
+    } else if (flags & 4) {
+        /* radix */
+        if (table_size > 24 || (proc_tbl & 0xfff) || (proc_tbl >> 60)) {
+            return H_PARAMETER;
+        }
+        cproc |= proc_tbl | table_size;
+    } else {
+        /* hash, possibly with process table */
+        if (table_size > 24 || (proc_tbl >> 38) || page_size > 7) {
+            return H_PARAMETER;
+        }
+        cproc = (proc_tbl << 25) | (page_size << 5) | table_size;
+    }
+    last_process_table = cproc;
+    fprintf(stderr, "calling config mmu flags=%lx proctbl=%lx\n",
+            cflags, cproc);
+    if  (!kvmppc_configure_v3_mmu(cpu, cflags, cproc)) {
+        return H_HARDWARE;
+    }
+    return H_SUCCESS;
+}
+
 static spapr_hcall_fn papr_hypercall_table[(MAX_HCALL_OPCODE / 4) + 1];
 static spapr_hcall_fn kvmppc_hypercall_table[KVMPPC_HCALL_MAX - KVMPPC_HCALL_BASE + 1];
 
@@ -1115,6 +1159,10 @@ static void hypercall_register_types(void)
 
     /* ibm,client-architecture-support support */
     spapr_register_hypercall(KVMPPC_H_CAS, h_client_architecture_support);
+
+    /* Power9 MMU support */
+    spapr_register_hypercall(H_REGISTER_PROC_TBL,
+                             h_register_process_table);
 }
 
 type_init(hypercall_register_types)
include/hw/ppc/spapr.h +1 -0
index a30cbc485c..d523db3b4a 100644
@@ -346,6 +346,7 @@ struct sPAPRMachineState {
 #define H_XIRR_X                0x2FC
 #define H_RANDOM                0x300
 #define H_SET_MODE              0x31C
+#define H_REGISTER_PROC_TBL     0x37C
 #define H_SIGNAL_SYS_RESET      0x380
 #define MAX_HCALL_OPCODE        H_SIGNAL_SYS_RESET
 
target/ppc/kvm.c +12 -0
index 8b153808fd..34dde45eef 100644
@@ -358,6 +358,18 @@ struct ppc_radix_page_info *kvm_get_radix_page_info(void)
     return radix_page_info;
 }
 
+bool kvmppc_configure_v3_mmu(PowerPCCPU *cpu, uint64_t flags, uint64_t proc_tbl)
+{
+    CPUState *cs = CPU(cpu);
+    int ret;
+    struct kvm_ppc_mmuv3_cfg cfg;
+
+    cfg.flags = flags;
+    cfg.process_table = proc_tbl;
+    ret = kvm_vm_ioctl(cs->kvm_state, KVM_PPC_CONFIGURE_V3_MMU, &cfg);
+    return ret == 0;
+}
+
 static long gethugepagesize(const char *mem_path)
 {
     struct statfs fs;
target/ppc/kvm_ppc.h +1 -0
index 56e222dfc2..441fa6a2db 100644
@@ -33,6 +33,7 @@ int kvmppc_clear_tsr_bits(PowerPCCPU *cpu, uint32_t tsr_bits);
 int kvmppc_or_tsr_bits(PowerPCCPU *cpu, uint32_t tsr_bits);
 int kvmppc_set_tcr(PowerPCCPU *cpu);
 int kvmppc_booke_watchdog_enable(PowerPCCPU *cpu);
+bool kvmppc_configure_v3_mmu(PowerPCCPU *cpu, uint64_t flags, uint64_t proctbl);
 #ifndef CONFIG_USER_ONLY
 off_t kvmppc_alloc_rma(void **rma);
 bool kvmppc_spapr_use_multitce(void);
Signature · 1 line
2.11.0
#23 lore raw

RFC PATCH v2 10/12 spapr: move spapr_populate_pa_features()

In reply to Sam Bobroff

In the next patch, spapr_fixup_cpu_dt() will need to call spapr_populate_pa_features() so move it's definition up without making any other changes.

---
 hw/ppc/spapr.c | 86 +++++++++++++++++++++++++++++-----------------------------
 1 file changed, 43 insertions(+), 43 deletions(-)
hw/ppc/spapr.c +43 -43
index dfee0f685f..0c0782b558 100644
@@ -194,6 +194,49 @@ static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
     return ret;
 }
 
+/* Populate the "ibm,pa-features" property */
+static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
+{
+    uint8_t pa_features_206[] = { 6, 0,
+        0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
+    uint8_t pa_features_207[] = { 24, 0,
+        0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
+        0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
+        0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
+        0x80, 0x00, 0x80, 0x00, 0x00, 0x00 };
+    uint8_t *pa_features;
+    size_t pa_size;
+
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
+    case POWERPC_MMU_VER_2_06:
+        pa_features = pa_features_206;
+        pa_size = sizeof(pa_features_206);
+        break;
+    case POWERPC_MMU_VER_2_07:
+        pa_features = pa_features_207;
+        pa_size = sizeof(pa_features_207);
+        break;
+    default:
+        return;
+    }
+
+    if (env->ci_large_pages) {
+        /*
+         * Note: we keep CI large pages off by default because a 64K capable
+         * guest provisioned with large pages might otherwise try to map a qemu
+         * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
+         * even if that qemu runs on a 4k host.
+         * We dd this bit back here if we are confident this is not an issue
+         */
+        pa_features[3] |= 0x20;
+    }
+    if (kvmppc_has_cap_htm() && pa_size > 24) {
+        pa_features[24] |= 0x80;    /* Transactional memory support */
+    }
+
+    _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
+}
+
 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
 {
     int ret = 0, offset, cpus_offset;
@@ -346,49 +389,6 @@ static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
     return 0;
 }
 
-/* Populate the "ibm,pa-features" property */
-static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
-{
-    uint8_t pa_features_206[] = { 6, 0,
-        0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
-    uint8_t pa_features_207[] = { 24, 0,
-        0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
-        0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
-        0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
-        0x80, 0x00, 0x80, 0x00, 0x00, 0x00 };
-    uint8_t *pa_features;
-    size_t pa_size;
-
-    switch (POWERPC_MMU_VER(env->mmu_model)) {
-    case POWERPC_MMU_VER_2_06:
-        pa_features = pa_features_206;
-        pa_size = sizeof(pa_features_206);
-        break;
-    case POWERPC_MMU_VER_2_07:
-        pa_features = pa_features_207;
-        pa_size = sizeof(pa_features_207);
-        break;
-    default:
-        return;
-    }
-
-    if (env->ci_large_pages) {
-        /*
-         * Note: we keep CI large pages off by default because a 64K capable
-         * guest provisioned with large pages might otherwise try to map a qemu
-         * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
-         * even if that qemu runs on a 4k host.
-         * We dd this bit back here if we are confident this is not an issue
-         */
-        pa_features[3] |= 0x20;
-    }
-    if (kvmppc_has_cap_htm() && pa_size > 24) {
-        pa_features[24] |= 0x80;    /* Transactional memory support */
-    }
-
-    _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
-}
-
 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
                                   sPAPRMachineState *spapr)
 {
Signature · 1 line
2.11.0
#24 lore raw

Re: [Qemu-devel] [RFC PATCH v2 10/12] spapr: move spapr_populate_pa_features()

Quoted 3 lines · Sam Bobroff

In the next patch, spapr_fixup_cpu_dt() will need to call spapr_populate_pa_features() so move it's definition up without making any other changes.

s/it's/its/

Quoted 110 lines
---
 hw/ppc/spapr.c | 86 +++++++++++++++++++++++++++++-----------------------------
 1 file changed, 43 insertions(+), 43 deletions(-)
hw/ppc/spapr.c +43 -43
index dfee0f685f..0c0782b558 100644
@@ -194,6 +194,49 @@ static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
     return ret;
 }
 
+/* Populate the "ibm,pa-features" property */
+static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
+{
+    uint8_t pa_features_206[] = { 6, 0,
+        0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
+    uint8_t pa_features_207[] = { 24, 0,
+        0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
+        0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
+        0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
+        0x80, 0x00, 0x80, 0x00, 0x00, 0x00 };
+    uint8_t *pa_features;
+    size_t pa_size;
+
+    switch (POWERPC_MMU_VER(env->mmu_model)) {
+    case POWERPC_MMU_VER_2_06:
+        pa_features = pa_features_206;
+        pa_size = sizeof(pa_features_206);
+        break;
+    case POWERPC_MMU_VER_2_07:
+        pa_features = pa_features_207;
+        pa_size = sizeof(pa_features_207);
+        break;
+    default:
+        return;
+    }
+
+    if (env->ci_large_pages) {
+        /*
+         * Note: we keep CI large pages off by default because a 64K capable
+         * guest provisioned with large pages might otherwise try to map a qemu
+         * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
+         * even if that qemu runs on a 4k host.
+         * We dd this bit back here if we are confident this is not an issue
+         */
+        pa_features[3] |= 0x20;
+    }
+    if (kvmppc_has_cap_htm() && pa_size > 24) {
+        pa_features[24] |= 0x80;    /* Transactional memory support */
+    }
+
+    _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
+}
+
 static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
 {
     int ret = 0, offset, cpus_offset;
@@ -346,49 +389,6 @@ static int spapr_populate_memory(sPAPRMachineState *spapr, void *fdt)
     return 0;
 }
 
-/* Populate the "ibm,pa-features" property */
-static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
-{
-    uint8_t pa_features_206[] = { 6, 0,
-        0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
-    uint8_t pa_features_207[] = { 24, 0,
-        0xf6, 0x1f, 0xc7, 0xc0, 0x80, 0xf0,
-        0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
-        0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
-        0x80, 0x00, 0x80, 0x00, 0x00, 0x00 };
-    uint8_t *pa_features;
-    size_t pa_size;
-
-    switch (POWERPC_MMU_VER(env->mmu_model)) {
-    case POWERPC_MMU_VER_2_06:
-        pa_features = pa_features_206;
-        pa_size = sizeof(pa_features_206);
-        break;
-    case POWERPC_MMU_VER_2_07:
-        pa_features = pa_features_207;
-        pa_size = sizeof(pa_features_207);
-        break;
-    default:
-        return;
-    }
-
-    if (env->ci_large_pages) {
-        /*
-         * Note: we keep CI large pages off by default because a 64K capable
-         * guest provisioned with large pages might otherwise try to map a qemu
-         * framebuffer (or other kind of memory mapped PCI BAR) using 64K pages
-         * even if that qemu runs on a 4k host.
-         * We dd this bit back here if we are confident this is not an issue
-         */
-        pa_features[3] |= 0x20;
-    }
-    if (kvmppc_has_cap_htm() && pa_size > 24) {
-        pa_features[24] |= 0x80;    /* Transactional memory support */
-    }
-
-    _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
-}
-
 static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
                                   sPAPRMachineState *spapr)
 {
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
#25 lore raw

RFC PATCH v2 11/12 spapr: Enable ISA 3.0 MMU mode selection via CAS

In reply to Sam Bobroff

Add the new node, /chosen/ibm,arch-vec-5-platform-support to the device tree. This allows the guest to determine which modes are supported by the hypervisor.

Update the option vector processing in h_client_architecture_support() to handle the new MMU bits. This allows guests to request hash or radix mode and QEMU to create the guest's HPT at this time if it is necessary but hasn't yet been done. QEMU will terminate the guest if it requests an unavailable mode, as required by the architecture.

Extend the ibm,pa-features node with the new ISA 3.0 values and set the radix bit if KVM supports radix mode. This probably won't be used directly by guests to determine the availability of radix mode (that is indicated by the new node added above) but the architecture requires that it be set when the hardware supports it.

ISA 3.0 guests will now begin to call h_register_process_table(), which has been added previously.

---
v2:

* Unused bits removed.
* Logic and bit definitions changed due to architectural change.
* Cleanly terminate QEMU if the guest requests an unavailable mode (as required by the new architecture).
* Legacy guest workaround moved to it's own patch.
* I'm sorry for the bitfield constants in spapr_dt_ov5_platform_support() but there don't seem to be convienent macros for converting an option vector specifier (OV_BIT(x,y)) into a byte-mask. I'm open to suggestions.

 hw/ppc/spapr.c              | 53 +++++++++++++++++++++++++++++++++++++++++++++
 hw/ppc/spapr_hcall.c        | 37 ++++++++++++++++++++++++-------
 include/hw/ppc/spapr_ovec.h |  5 +++++
 3 files changed, 87 insertions(+), 8 deletions(-)
hw/ppc/spapr.c +53 -0
index 0c0782b558..e83468a8d3 100644
@@ -204,6 +204,20 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
         0x80, 0x00, 0x00, 0x00, 0x00, 0x00,
         0x00, 0x00, 0x00, 0x00, 0x80, 0x00,
         0x80, 0x00, 0x80, 0x00, 0x00, 0x00 };
+    uint8_t pa_features_300[70 + 2] = { 70, 0,
+        0xf6, 0x3f, 0xc7, 0xc0, 0x80, 0xf0, /* 0 - 5 */
+        0x80, 0x00, 0x00, 0x00, 0x00, 0x00, /* 6 - 11 */
+        0x00, 0x00, 0x00, 0x00, 0x80, 0x00, /* 12 - 17 */
+        0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 18 - 23 */
+        0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 24 - 29 */
+        0x80, 0x00, 0x80, 0x00, 0xC0, 0x00, /* 30 - 35 */
+        0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 36 - 41 */
+        0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 42 - 47 */
+        0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 48 - 53 */
+        0x80, 0x00, 0x80, 0x00, 0x80, 0x00, /* 54 - 59 */
+        0x80, 0x00, 0x80, 0x00, 0x00, 0x00, /* 60 - 64 */
+        0x00, 0x00, 0x00, 0x00,             /* 66 - 69 */
+        };
     uint8_t *pa_features;
     size_t pa_size;
 
@@ -216,6 +230,10 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
         pa_features = pa_features_207;
         pa_size = sizeof(pa_features_207);
         break;
+    case POWERPC_MMU_VER_3_00:
+        pa_features = pa_features_300;
+        pa_size = sizeof(pa_features_300);
+        break;
     default:
         return;
     }
@@ -804,6 +822,34 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
     spapr_dt_rtas_tokens(fdt, rtas);
 }
 
+/* Prepare ibm,arch-vec-5-platform-support, which indicates the MMU features
+ * that the guest may request and thus the valid values for bytes 24..26 of
+ * option vector 5: */
+static void spapr_dt_ov5_platform_support(void *fdt, int chosen)
+{
+    char val[2 * 3] = {
+        24, 0x00, /* Hash/Radix, filled in below. */
+        25, 0x40, /* Hash options: Segment Tables == no, GTSE == no. */
+        26, 0x40, /* Radix options: GTSE == yes. */
+    };
+
+    if (kvm_enabled()) {
+        if (kvmppc_has_cap_mmu_radix() && kvmppc_has_cap_mmu_hash_v3()) {
+            val[1] = 0x80; /* OV5_MMU_BOTH */
+        } else if (kvmppc_has_cap_mmu_radix()) {
+            val[1] = 0x40; /* OV5_MMU_RADIX_300 */
+        } else {
+            assert(kvmppc_has_cap_mmu_hash_v3());
+            val[1] = 0x00; /* Hash */
+        }
+    } else {
+        /* TODO: TCG case, hash */
+        val[1] = 0x00;
+    }
+    _FDT(fdt_setprop(fdt, chosen, "ibm,arch-vec-5-platform-support",
+                     val, sizeof(val)));
+}
+
 static void spapr_dt_chosen(sPAPRMachineState *spapr, void *fdt)
 {
     MachineState *machine = MACHINE(spapr);
@@ -857,6 +903,8 @@ static void spapr_dt_chosen(sPAPRMachineState *spapr, void *fdt)
         _FDT(fdt_setprop_string(fdt, chosen, "linux,stdout-path", stdout_path));
     }
 
+    spapr_dt_ov5_platform_support(fdt, chosen);
+
     g_free(stdout_path);
     g_free(bootlist);
 }
@@ -1929,6 +1977,11 @@ static void ppc_spapr_init(MachineState *machine)
     }
 
     spapr_ovec_set(spapr->ov5, OV5_FORM1_AFFINITY);
+    if (kvmppc_has_cap_mmu_radix()) {
+        /* KVM always allows GTSE with radix... */
+        spapr_ovec_set(spapr->ov5, OV5_MMU_RADIX_GTSE);
+    }
+    /* ... but not with hash (currently). */
 
     /* advertise support for dedicated HP event source to guests */
     if (spapr->use_hotplug_event_source) {
hw/ppc/spapr_hcall.c +29 -8
index 9391619ed6..efaa1a1b19 100644
@@ -12,6 +12,7 @@
 #include "trace.h"
 #include "kvm_ppc.h"
 #include "hw/ppc/spapr_ovec.h"
+#include "qemu/error-report.h"
 
 struct SPRSyncState {
     int spr;
@@ -933,6 +934,7 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     uint32_t best_compat = 0;
     int i;
     sPAPROptionVector *ov5_guest, *ov5_cas_old, *ov5_updates;
+    bool guest_radix;
 
     /*
      * We scan the supplied table of PVRs looking for two things
@@ -984,6 +986,13 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     ov_table = list;
 
     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
+    if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
+        error_report("qemu: guest requested hash and radix MMU, which is invalid.");
+        exit(EXIT_FAILURE);
+    }
+    /* The radix/hash bit in byte 24 requires special handling: */
+    guest_radix = spapr_ovec_test(ov5_guest, OV5_MMU_RADIX_300);
+    spapr_ovec_clear(ov5_guest, OV5_MMU_RADIX_300);
 
     /* NOTE: there are actually a number of ov5 bits where input from the
      * guest is always zero, and the platform/QEMU enables them independently
@@ -1002,14 +1011,18 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     ov5_updates = spapr_ovec_new();
     spapr->cas_reboot = spapr_ovec_diff(ov5_updates,
                                         ov5_cas_old, spapr->ov5_cas);
-    if (kvm_enabled()) {
-        if (kvmppc_has_cap_mmu_radix()) {
-            /* If the HPT hasn't yet been set up (see
-             * ppc_spapr_reset()), and it's needed, do it now: */
-            if (!spapr_ovec_test(ov5_updates, OV5_MMU_RADIX)) {
-                /* legacy hash or new hash: */
-                spapr_setup_hpt_and_vrma(spapr);
-            }
+    /* Now that processing is finished, set the radix/hash bit for the
+     * guest if it requested a valid mode; otherwise terminate the boot. */
+    if (guest_radix) {
+        if (kvm_enabled() && !kvmppc_has_cap_mmu_radix()) {
+            error_report("qemu: Guest requested radix MMU mode when it is not available.");
+            exit(EXIT_FAILURE);
+        }
+        spapr_ovec_set(spapr->ov5_cas, OV5_MMU_RADIX_300);
+    } else {
+        if (kvm_enabled() && !kvmppc_has_cap_mmu_hash_v3()) {
+            error_report("qemu: Guest requested hash MMU mode when it is not available.");
+            exit(EXIT_FAILURE);
         }
     }
 
@@ -1022,6 +1035,14 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
 
     if (spapr->cas_reboot) {
         qemu_system_reset_request();
+    } else {
+        /* If ppc_spapr_reset() did not set up a HPT but one is necessary
+         * (because the guest isn't going to use radix) then set it up here. */
+        if (kvm_enabled()) {
+            if (kvmppc_has_cap_mmu_radix() && !guest_radix) {
+                spapr_setup_hpt_and_vrma(spapr);
+            }
+        }
     }
 
     return H_SUCCESS;
include/hw/ppc/spapr_ovec.h +5 -0
index 355a34411f..e2dfbac558 100644
@@ -48,6 +48,11 @@ typedef struct sPAPROptionVector sPAPROptionVector;
 #define OV5_FORM1_AFFINITY      OV_BIT(5, 0)
 #define OV5_HP_EVT              OV_BIT(6, 5)
 
+/* ISA 3.00 MMU features: */
+#define OV5_MMU_BOTH            OV_BIT(24, 0) /* Radix and hash */
+#define OV5_MMU_RADIX_300       OV_BIT(24, 1) /* 1 => Radix only, 0 => Hash only */
+#define OV5_MMU_RADIX_GTSE      OV_BIT(26, 1) /* Radix GTSE */
+
 /* interfaces */
 sPAPROptionVector *spapr_ovec_new(void);
 sPAPROptionVector *spapr_ovec_clone(sPAPROptionVector *ov_orig);
Signature · 1 line
2.11.0
#26 lore raw

RFC PATCH v2 12/12 spapr: Workaround for broken radix guests

In reply to Sam Bobroff

For a little while around 4.9, Linux kernels that saw the radix bit in ibm,pa-features would attempt to set up the MMU as if they were a hypervisor, even if they were a guest, which would cause them to crash.

Work around this by detecting pre-ISA 3.0 guests by their lack of that bit in option vector 1, and then removing the radix bit from ibm,pa-features. Note: This now requires regeneration of that node after CAS negotiation.

---
 hw/ppc/spapr.c              | 15 +++++++++++++--
 hw/ppc/spapr_hcall.c        |  5 +++--
 include/hw/ppc/spapr.h      |  1 +
 include/hw/ppc/spapr_ovec.h |  3 +++
 4 files changed, 20 insertions(+), 4 deletions(-)
hw/ppc/spapr.c +13 -2
index e83468a8d3..c47600b8ee 100644
@@ -195,7 +195,8 @@ static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
 }
 
 /* Populate the "ibm,pa-features" property */
-static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
+static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset,
+                                      bool legacy_guest)
 {
     uint8_t pa_features_206[] = { 6, 0,
         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
@@ -251,6 +252,12 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
     if (kvmppc_has_cap_htm() && pa_size > 24) {
         pa_features[24] |= 0x80;    /* Transactional memory support */
     }
+    if (legacy_guest && pa_size > 40) {
+        /* Workaround for broken kernels that attempt (guest) radix
+         * mode when they can't handle it, if they see the radix bit set
+         * in pa-features. So hide it from them. */
+        pa_features[40 + 2] &= ~0x80; /* Radix MMU */
+    }
 
     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
 }
@@ -265,6 +272,7 @@ static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
 
     CPU_FOREACH(cs) {
         PowerPCCPU *cpu = POWERPC_CPU(cs);
+        CPUPPCState *env = &cpu->env;
         DeviceClass *dc = DEVICE_GET_CLASS(cs);
         int index = ppc_get_vcpu_dt_id(cpu);
         int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu));
@@ -306,6 +314,9 @@ static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
         if (ret < 0) {
             return ret;
         }
+
+        spapr_populate_pa_features(env, fdt, offset,
+                                         spapr->cas_legacy_guest_workaround);
     }
     return ret;
 }
@@ -503,7 +514,7 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
                           page_sizes_prop, page_sizes_prop_size)));
     }
 
-    spapr_populate_pa_features(env, fdt, offset);
+    spapr_populate_pa_features(env, fdt, offset, false);
 
     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
                            cs->cpu_index / vcpus_per_socket)));
hw/ppc/spapr_hcall.c +3 -2
index efaa1a1b19..7660cd7d64 100644
@@ -933,7 +933,7 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     uint32_t max_compat = cpu->max_compat;
     uint32_t best_compat = 0;
     int i;
-    sPAPROptionVector *ov5_guest, *ov5_cas_old, *ov5_updates;
+    sPAPROptionVector *ov1_guest, *ov5_guest, *ov5_cas_old, *ov5_updates;
     bool guest_radix;
 
     /*
@@ -985,6 +985,7 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     /* For the future use: here @ov_table points to the first option vector */
     ov_table = list;
 
+    ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
     if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
         error_report("qemu: guest requested hash and radix MMU, which is invalid.");
@@ -1025,7 +1026,7 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
             exit(EXIT_FAILURE);
         }
     }
-
+    spapr->cas_legacy_guest_workaround = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
     if (!spapr->cas_reboot) {
         spapr->cas_reboot =
             (spapr_h_cas_compose_response(spapr, args[1], args[2],
include/hw/ppc/spapr.h +1 -0
index d523db3b4a..1e64e3ada8 100644
@@ -77,6 +77,7 @@ struct sPAPRMachineState {
     sPAPROptionVector *ov5;         /* QEMU-supported option vectors */
     sPAPROptionVector *ov5_cas;     /* negotiated (via CAS) option vectors */
     bool cas_reboot;
+    bool cas_legacy_guest_workaround;
 
     Notifier epow_notifier;
     QTAILQ_HEAD(, sPAPREventLogEntry) pending_events;
include/hw/ppc/spapr_ovec.h +3 -0
index e2dfbac558..8807c753e0 100644
@@ -43,6 +43,9 @@ typedef struct sPAPROptionVector sPAPROptionVector;
 
 #define OV_BIT(byte, bit) ((byte - 1) * BITS_PER_BYTE + bit)
 
+/* option vector 1 */
+#define OV1_PPC_3_00            OV_BIT(3, 0) /* set if we support PowerPC 3.00 */
+
 /* option vector 5 */
 #define OV5_DRCONF_MEMORY       OV_BIT(2, 2)
 #define OV5_FORM1_AFFINITY      OV_BIT(5, 0)
Signature · 1 line
2.11.0
#27 lore raw

Re: [Qemu-devel] [RFC PATCH v2 12/12] spapr: Workaround for broken radix guests

Quoted 11 lines · Sam Bobroff

For a little while around 4.9, Linux kernels that saw the radix bit in ibm,pa-features would attempt to set up the MMU as if they were a hypervisor, even if they were a guest, which would cause them to crash.

Work around this by detecting pre-ISA 3.0 guests by their lack of that bit in option vector 1, and then removing the radix bit from ibm,pa-features. Note: This now requires regeneration of that node after CAS negotiation.

A bit ugly, but not any more so than it needs to given what we're dealing with AFAICT.

I'll save more detailed review until the rebase in conjuction with the TCG bits.

Quoted 117 lines
---
 hw/ppc/spapr.c              | 15 +++++++++++++--
 hw/ppc/spapr_hcall.c        |  5 +++--
 include/hw/ppc/spapr.h      |  1 +
 include/hw/ppc/spapr_ovec.h |  3 +++
 4 files changed, 20 insertions(+), 4 deletions(-)
hw/ppc/spapr.c +13 -2
index e83468a8d3..c47600b8ee 100644
@@ -195,7 +195,8 @@ static int spapr_fixup_cpu_numa_dt(void *fdt, int offset, CPUState *cs)
 }
 
 /* Populate the "ibm,pa-features" property */
-static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
+static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset,
+                                      bool legacy_guest)
 {
     uint8_t pa_features_206[] = { 6, 0,
         0xf6, 0x1f, 0xc7, 0x00, 0x80, 0xc0 };
@@ -251,6 +252,12 @@ static void spapr_populate_pa_features(CPUPPCState *env, void *fdt, int offset)
     if (kvmppc_has_cap_htm() && pa_size > 24) {
         pa_features[24] |= 0x80;    /* Transactional memory support */
     }
+    if (legacy_guest && pa_size > 40) {
+        /* Workaround for broken kernels that attempt (guest) radix
+         * mode when they can't handle it, if they see the radix bit set
+         * in pa-features. So hide it from them. */
+        pa_features[40 + 2] &= ~0x80; /* Radix MMU */
+    }
 
     _FDT((fdt_setprop(fdt, offset, "ibm,pa-features", pa_features, pa_size)));
 }
@@ -265,6 +272,7 @@ static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
 
     CPU_FOREACH(cs) {
         PowerPCCPU *cpu = POWERPC_CPU(cs);
+        CPUPPCState *env = &cpu->env;
         DeviceClass *dc = DEVICE_GET_CLASS(cs);
         int index = ppc_get_vcpu_dt_id(cpu);
         int compat_smt = MIN(smp_threads, ppc_compat_max_threads(cpu));
@@ -306,6 +314,9 @@ static int spapr_fixup_cpu_dt(void *fdt, sPAPRMachineState *spapr)
         if (ret < 0) {
             return ret;
         }
+
+        spapr_populate_pa_features(env, fdt, offset,
+                                         spapr->cas_legacy_guest_workaround);
     }
     return ret;
 }
@@ -503,7 +514,7 @@ static void spapr_populate_cpu_dt(CPUState *cs, void *fdt, int offset,
                           page_sizes_prop, page_sizes_prop_size)));
     }
 
-    spapr_populate_pa_features(env, fdt, offset);
+    spapr_populate_pa_features(env, fdt, offset, false);
 
     _FDT((fdt_setprop_cell(fdt, offset, "ibm,chip-id",
                            cs->cpu_index / vcpus_per_socket)));
hw/ppc/spapr_hcall.c +3 -2
index efaa1a1b19..7660cd7d64 100644
@@ -933,7 +933,7 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     uint32_t max_compat = cpu->max_compat;
     uint32_t best_compat = 0;
     int i;
-    sPAPROptionVector *ov5_guest, *ov5_cas_old, *ov5_updates;
+    sPAPROptionVector *ov1_guest, *ov5_guest, *ov5_cas_old, *ov5_updates;
     bool guest_radix;
 
     /*
@@ -985,6 +985,7 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
     /* For the future use: here @ov_table points to the first option vector */
     ov_table = list;
 
+    ov1_guest = spapr_ovec_parse_vector(ov_table, 1);
     ov5_guest = spapr_ovec_parse_vector(ov_table, 5);
     if (spapr_ovec_test(ov5_guest, OV5_MMU_BOTH)) {
         error_report("qemu: guest requested hash and radix MMU, which is invalid.");
@@ -1025,7 +1026,7 @@ static target_ulong h_client_architecture_support(PowerPCCPU *cpu,
             exit(EXIT_FAILURE);
         }
     }
-
+    spapr->cas_legacy_guest_workaround = !spapr_ovec_test(ov1_guest, OV1_PPC_3_00);
     if (!spapr->cas_reboot) {
         spapr->cas_reboot =
             (spapr_h_cas_compose_response(spapr, args[1], args[2],
include/hw/ppc/spapr.h +1 -0
index d523db3b4a..1e64e3ada8 100644
@@ -77,6 +77,7 @@ struct sPAPRMachineState {
     sPAPROptionVector *ov5;         /* QEMU-supported option vectors */
     sPAPROptionVector *ov5_cas;     /* negotiated (via CAS) option vectors */
     bool cas_reboot;
+    bool cas_legacy_guest_workaround;
 
     Notifier epow_notifier;
     QTAILQ_HEAD(, sPAPREventLogEntry) pending_events;
include/hw/ppc/spapr_ovec.h +3 -0
index e2dfbac558..8807c753e0 100644
@@ -43,6 +43,9 @@ typedef struct sPAPROptionVector sPAPROptionVector;
 
 #define OV_BIT(byte, bit) ((byte - 1) * BITS_PER_BYTE + bit)
 
+/* option vector 1 */
+#define OV1_PPC_3_00            OV_BIT(3, 0) /* set if we support PowerPC 3.00 */
+
 /* option vector 5 */
 #define OV5_DRCONF_MEMORY       OV_BIT(2, 2)
 #define OV5_FORM1_AFFINITY      OV_BIT(5, 0)
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