hax: Support guest RAM sizes of 4GB or more

Since HAX_VM_IOCTL_ALLOC_RAM takes a 32-bit size, it cannot handle RAM blocks of 4GB or larger, which is why HAXM can only run guests with less than 4GB of RAM. Solve this problem by utilizing the new HAXM API, HAX_VM_IOCTL_ADD_RAMBLOCK, which takes a 64-bit size, to register RAM blocks with the HAXM kernel module. The new API is first added in HAXM 7.0.0, and its availablility and be confirmed by the presence of the HAX_CAP_64BIT_RAMBLOCK capability flag. When the guest RAM size reaches 7GB, QEMU will ask HAXM to set up a memory mapping that covers a 4GB region, which will fail, because HAX_VM_IOCTL_SET_RAM also takes a 32-bit size. Work around this limitation by splitting the large mapping into small ones and calling HAX_VM_IOCTL_SET_RAM multiple times. Bug: https://bugs.launchpad.net/qemu/+bug/1735576 Signed-off-by: Yu Ning <[email protected]> Message-Id: <[email protected]> Signed-off-by: Paolo Bonzini <[email protected]>
n64decomp · Feb 13, 2018 · 7a5235c · 7a5235c
1 parent 7b40951
commit 7a5235c
Show file tree

Hide file tree

Showing 9 changed files with 89 additions and 26 deletions.
diff --git a/include/sysemu/hax.h b/include/sysemu/hax.h
@@ -27,7 +27,7 @@
 int hax_sync_vcpus(void);
 int hax_init_vcpu(CPUState *cpu);
 int hax_smp_cpu_exec(CPUState *cpu);
-int hax_populate_ram(uint64_t va, uint32_t size);
+int hax_populate_ram(uint64_t va, uint64_t size);
 
 void hax_cpu_synchronize_state(CPUState *cpu);
 void hax_cpu_synchronize_post_reset(CPUState *cpu);

diff --git a/target/i386/hax-all.c b/target/i386/hax-all.c
@@ -103,6 +103,8 @@ static int hax_get_capability(struct hax_state *hax)
         return -ENOTSUP;
     }
 
+    hax->supports_64bit_ramblock = !!(cap->winfo & HAX_CAP_64BIT_RAMBLOCK);
+
     if (cap->wstatus & HAX_CAP_MEMQUOTA) {
         if (cap->mem_quota < hax->mem_quota) {
             fprintf(stderr, "The VM memory needed exceeds the driver limit.\n");

diff --git a/target/i386/hax-darwin.c b/target/i386/hax-darwin.c
@@ -28,21 +28,36 @@ hax_fd hax_mod_open(void)
     return fd;
 }
 
-int hax_populate_ram(uint64_t va, uint32_t size)
+int hax_populate_ram(uint64_t va, uint64_t size)
 {
     int ret;
-    struct hax_alloc_ram_info info;
 
     if (!hax_global.vm || !hax_global.vm->fd) {
         fprintf(stderr, "Allocate memory before vm create?\n");
         return -EINVAL;
     }
 
-    info.size = size;
-    info.va = va;
-    ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_ALLOC_RAM, &info);
+    if (hax_global.supports_64bit_ramblock) {
+        struct hax_ramblock_info ramblock = {
+            .start_va = va,
+            .size = size,
+            .reserved = 0
+        };
+
+        ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_ADD_RAMBLOCK, &ramblock);
+    } else {
+        struct hax_alloc_ram_info info = {
+            .size = (uint32_t)size,
+            .pad = 0,
+            .va = va
+        };
+
+        ret = ioctl(hax_global.vm->fd, HAX_VM_IOCTL_ALLOC_RAM, &info);
+    }
     if (ret < 0) {
-        fprintf(stderr, "Failed to allocate %x memory\n", size);
+        fprintf(stderr, "Failed to register RAM block: ret=%d, va=0x%" PRIx64
+                ", size=0x%" PRIx64 ", method=%s\n", ret, va, size,
+                hax_global.supports_64bit_ramblock ? "new" : "legacy");
         return ret;
     }
     return 0;

diff --git a/target/i386/hax-darwin.h b/target/i386/hax-darwin.h
@@ -44,6 +44,7 @@ static inline void hax_close_fd(hax_fd fd)
 #define HAX_VM_IOCTL_SET_RAM _IOWR(0, 0x82, struct hax_set_ram_info)
 #define HAX_VM_IOCTL_VCPU_DESTROY _IOW(0, 0x83, uint32_t)
 #define HAX_VM_IOCTL_NOTIFY_QEMU_VERSION _IOW(0, 0x84, struct hax_qemu_version)
+#define HAX_VM_IOCTL_ADD_RAMBLOCK _IOW(0, 0x85, struct hax_ramblock_info)
 
 #define HAX_VCPU_IOCTL_RUN  _IO(0, 0xc0)
 #define HAX_VCPU_IOCTL_SET_MSRS _IOWR(0, 0xc1, struct hax_msr_data)

diff --git a/target/i386/hax-i386.h b/target/i386/hax-i386.h
@@ -37,6 +37,7 @@ struct hax_state {
     uint32_t version;
     struct hax_vm *vm;
     uint64_t mem_quota;
+    bool supports_64bit_ramblock;
 };
 
 #define HAX_MAX_VCPU 0x10

diff --git a/target/i386/hax-interface.h b/target/i386/hax-interface.h
@@ -308,6 +308,13 @@ struct hax_alloc_ram_info {
     uint32_t pad;
     uint64_t va;
 } __attribute__ ((__packed__));
+
+struct hax_ramblock_info {
+    uint64_t start_va;
+    uint64_t size;
+    uint64_t reserved;
+} __attribute__ ((__packed__));
+
 #define HAX_RAM_INFO_ROM     0x01 /* Read-Only */
 #define HAX_RAM_INFO_INVALID 0x80 /* Unmapped, usually used for MMIO */
 struct hax_set_ram_info {
@@ -327,6 +334,7 @@ struct hax_set_ram_info {
 
 #define HAX_CAP_MEMQUOTA           0x2
 #define HAX_CAP_UG                 0x4
+#define HAX_CAP_64BIT_RAMBLOCK     0x8
 
 struct hax_capabilityinfo {
     /* bit 0: 1 - working

diff --git a/target/i386/hax-mem.c b/target/i386/hax-mem.c
@@ -174,6 +174,7 @@ static void hax_process_section(MemoryRegionSection *section, uint8_t flags)
     ram_addr_t size = int128_get64(section->size);
     unsigned int delta;
     uint64_t host_va;
+    uint32_t max_mapping_size;
 
     /* We only care about RAM and ROM regions */
     if (!memory_region_is_ram(mr)) {
@@ -206,10 +207,23 @@ static void hax_process_section(MemoryRegionSection *section, uint8_t flags)
         flags |= HAX_RAM_INFO_ROM;
     }
 
-    /* the kernel module interface uses 32-bit sizes (but we could split...) */
-    g_assert(size <= UINT32_MAX);
-
-    hax_update_mapping(start_pa, size, host_va, flags);
+    /*
+     * The kernel module interface uses 32-bit sizes:
+     * https://github.com/intel/haxm/blob/master/API.md#hax_vm_ioctl_set_ram
+     *
+     * If the mapping size is longer than 32 bits, we can't process it in one
+     * call into the kernel. Instead, we split the mapping into smaller ones,
+     * and call hax_update_mapping() on each.
+     */
+    max_mapping_size = UINT32_MAX & qemu_real_host_page_mask;
+    while (size > max_mapping_size) {
+        hax_update_mapping(start_pa, max_mapping_size, host_va, flags);
+        start_pa += max_mapping_size;
+        size -= max_mapping_size;
+        host_va += max_mapping_size;
+    }
+    /* Now size <= max_mapping_size */
+    hax_update_mapping(start_pa, (uint32_t)size, host_va, flags);
 }
 
 static void hax_region_add(MemoryListener *listener,
@@ -283,12 +297,16 @@ static MemoryListener hax_memory_listener = {
 static void hax_ram_block_added(RAMBlockNotifier *n, void *host, size_t size)
 {
     /*
-     * In HAX, QEMU allocates the virtual address, and HAX kernel
-     * populates the memory with physical memory. Currently we have no
-     * paging, so user should make sure enough free memory in advance.
+     * We must register each RAM block with the HAXM kernel module, or
+     * hax_set_ram() will fail for any mapping into the RAM block:
+     * https://github.com/intel/haxm/blob/master/API.md#hax_vm_ioctl_alloc_ram
+     *
+     * Old versions of the HAXM kernel module (< 6.2.0) used to preallocate all
+     * host physical pages for the RAM block as part of this registration
+     * process, hence the name hax_populate_ram().
      */
     if (hax_populate_ram((uint64_t)(uintptr_t)host, size) < 0) {
-        fprintf(stderr, "HAX failed to populate RAM");
+        fprintf(stderr, "HAX failed to populate RAM\n");
         abort();
     }
 }

diff --git a/target/i386/hax-windows.c b/target/i386/hax-windows.c
@@ -58,10 +58,9 @@ static int hax_open_device(hax_fd *fd)
     return fd;
 }
 
-int hax_populate_ram(uint64_t va, uint32_t size)
+int hax_populate_ram(uint64_t va, uint64_t size)
 {
     int ret;
-    struct hax_alloc_ram_info info;
     HANDLE hDeviceVM;
     DWORD dSize = 0;
 
@@ -70,18 +69,35 @@ int hax_populate_ram(uint64_t va, uint32_t size)
         return -EINVAL;
     }
 
-    info.size = size;
-    info.va = va;
-
     hDeviceVM = hax_global.vm->fd;
-
-    ret = DeviceIoControl(hDeviceVM,
-                          HAX_VM_IOCTL_ALLOC_RAM,
-                          &info, sizeof(info), NULL, 0, &dSize,
-                          (LPOVERLAPPED) NULL);
+    if (hax_global.supports_64bit_ramblock) {
+        struct hax_ramblock_info ramblock = {
+            .start_va = va,
+            .size = size,
+            .reserved = 0
+        };
+
+        ret = DeviceIoControl(hDeviceVM,
+                              HAX_VM_IOCTL_ADD_RAMBLOCK,
+                              &ramblock, sizeof(ramblock), NULL, 0, &dSize,
+                              (LPOVERLAPPED) NULL);
+    } else {
+        struct hax_alloc_ram_info info = {
+            .size = (uint32_t) size,
+            .pad = 0,
+            .va = va
+        };
+
+        ret = DeviceIoControl(hDeviceVM,
+                              HAX_VM_IOCTL_ALLOC_RAM,
+                              &info, sizeof(info), NULL, 0, &dSize,
+                              (LPOVERLAPPED) NULL);
+    }
 
     if (!ret) {
-        fprintf(stderr, "Failed to allocate %x memory\n", size);
+        fprintf(stderr, "Failed to register RAM block: va=0x%" PRIx64
+                ", size=0x%" PRIx64 ", method=%s\n", va, size,
+                hax_global.supports_64bit_ramblock ? "new" : "legacy");
         return ret;
     }
 

diff --git a/target/i386/hax-windows.h b/target/i386/hax-windows.h
@@ -57,6 +57,8 @@ static inline int hax_invalid_fd(hax_fd fd)
                                             METHOD_BUFFERED, FILE_ANY_ACCESS)
 #define HAX_VM_IOCTL_VCPU_DESTROY  CTL_CODE(HAX_DEVICE_TYPE, 0x905, \
                                             METHOD_BUFFERED, FILE_ANY_ACCESS)
+#define HAX_VM_IOCTL_ADD_RAMBLOCK  CTL_CODE(HAX_DEVICE_TYPE, 0x913, \
+                                            METHOD_BUFFERED, FILE_ANY_ACCESS)
 
 #define HAX_VCPU_IOCTL_RUN      CTL_CODE(HAX_DEVICE_TYPE, 0x906, \
                                          METHOD_BUFFERED, FILE_ANY_ACCESS)