[Qemu-devel] [7088] Implement dynamic guest ram allocation.

[Paul Brook]

Revision: 7088
          http://svn.sv.gnu.org/viewvc/?view=rev&root=qemu&revision=7088
Author:   pbrook
Date:     2009-04-11 17:15:54 +0000 (Sat, 11 Apr 2009)
Log Message:
-----------
Implement dynamic guest ram allocation.

Signed-off-by: Paul Brook <address@hidden>

Modified Paths:
--------------
    trunk/cpu-all.h
    trunk/exec-all.h
    trunk/exec.c
    trunk/kqemu.c
    trunk/vl.c

Modified: trunk/cpu-all.h
===================================================================
--- trunk/cpu-all.h     2009-04-11 17:01:42 UTC (rev 7087)
+++ trunk/cpu-all.h     2009-04-11 17:15:54 UTC (rev 7088)
@@ -854,11 +854,10 @@
 
 /* memory API */
 
-extern ram_addr_t phys_ram_size;
 extern int phys_ram_fd;
-extern uint8_t *phys_ram_base;
 extern uint8_t *phys_ram_dirty;
 extern ram_addr_t ram_size;
+extern ram_addr_t last_ram_offset;
 
 /* physical memory access */
 

Modified: trunk/exec-all.h
===================================================================
--- trunk/exec-all.h    2009-04-11 17:01:42 UTC (rev 7087)
+++ trunk/exec-all.h    2009-04-11 17:15:54 UTC (rev 7088)
@@ -370,6 +370,9 @@
 
 extern uint32_t kqemu_comm_base;
 
+extern ram_addr_t kqemu_phys_ram_size;
+extern uint8_t *kqemu_phys_ram_base;
+
 static inline int kqemu_is_ok(CPUState *env)
 {
     return(env->kqemu_enabled &&

Modified: trunk/exec.c
===================================================================
--- trunk/exec.c        2009-04-11 17:01:42 UTC (rev 7087)
+++ trunk/exec.c        2009-04-11 17:15:54 UTC (rev 7088)
@@ -107,12 +107,22 @@
 uint8_t *code_gen_ptr;
 
 #if !defined(CONFIG_USER_ONLY)
-ram_addr_t phys_ram_size;
 int phys_ram_fd;
-uint8_t *phys_ram_base;
 uint8_t *phys_ram_dirty;
 static int in_migration;
-static ram_addr_t phys_ram_alloc_offset = 0;
+
+typedef struct RAMBlock {
+    uint8_t *host;
+    ram_addr_t offset;
+    ram_addr_t length;
+    struct RAMBlock *next;
+} RAMBlock;
+
+static RAMBlock *ram_blocks;
+/* TODO: When we implement (and use) ram deallocation (e.g. for hotplug)
+   then we can no longet assume contiguous ram offsets, and external uses
+   of this variable will break.  */
+ram_addr_t last_ram_offset;
 #endif
 
 CPUState *first_cpu;
@@ -411,7 +421,7 @@
         code_gen_buffer_size = DEFAULT_CODE_GEN_BUFFER_SIZE;
 #else
         /* XXX: needs ajustments */
-        code_gen_buffer_size = (unsigned long)(phys_ram_size / 4);
+        code_gen_buffer_size = (unsigned long)(ram_size / 4);
 #endif
     }
     if (code_gen_buffer_size < MIN_CODE_GEN_BUFFER_SIZE)
@@ -2419,22 +2429,55 @@
         kvm_uncoalesce_mmio_region(addr, size);
 }
 
+#ifdef USE_KQEMU
 /* XXX: better than nothing */
-ram_addr_t qemu_ram_alloc(ram_addr_t size)
+static ram_addr_t kqemu_ram_alloc(ram_addr_t size)
 {
     ram_addr_t addr;
-    if ((phys_ram_alloc_offset + size) > phys_ram_size) {
+    if ((last_ram_offset + size) > kqemu_phys_ram_size) {
         fprintf(stderr, "Not enough memory (requested_size = %" PRIu64 ", max 
memory = %" PRIu64 ")\n",
-                (uint64_t)size, (uint64_t)phys_ram_size);
+                (uint64_t)size, (uint64_t)kqemu_phys_ram_size);
         abort();
     }
-    addr = phys_ram_alloc_offset;
-    phys_ram_alloc_offset = TARGET_PAGE_ALIGN(phys_ram_alloc_offset + size);
+    addr = last_ram_offset;
+    last_ram_offset = TARGET_PAGE_ALIGN(last_ram_offset + size);
     return addr;
 }
+#endif
 
+ram_addr_t qemu_ram_alloc(ram_addr_t size)
+{
+    RAMBlock *new_block;
+
+#ifdef USE_KQEMU
+    if (kqemu_phys_ram_base) {
+        return kqemu_ram_alloc(size);
+    }
+#endif
+
+    size = TARGET_PAGE_ALIGN(size);
+    new_block = qemu_malloc(sizeof(*new_block));
+
+    new_block->host = qemu_vmalloc(size);
+    new_block->offset = last_ram_offset;
+    new_block->length = size;
+
+    new_block->next = ram_blocks;
+    ram_blocks = new_block;
+
+    phys_ram_dirty = qemu_realloc(phys_ram_dirty,
+        (last_ram_offset + size) >> TARGET_PAGE_BITS);
+    memset(phys_ram_dirty + (last_ram_offset >> TARGET_PAGE_BITS),
+           0xff, size >> TARGET_PAGE_BITS);
+
+    last_ram_offset += size;
+
+    return new_block->offset;
+}
+
 void qemu_ram_free(ram_addr_t addr)
 {
+    /* TODO: implement this.  */
 }
 
 /* Return a host pointer to ram allocated with qemu_ram_alloc.
@@ -2447,14 +2490,69 @@
  */
 void *qemu_get_ram_ptr(ram_addr_t addr)
 {
-    return phys_ram_base + addr;
+    RAMBlock *prev;
+    RAMBlock **prevp;
+    RAMBlock *block;
+
+#ifdef USE_KQEMU
+    if (kqemu_phys_ram_base) {
+        return kqemu_phys_ram_base + addr;
+    }
+#endif
+
+    prev = NULL;
+    prevp = &ram_blocks;
+    block = ram_blocks;
+    while (block && (block->offset > addr
+                     || block->offset + block->length <= addr)) {
+        if (prev)
+          prevp = &prev->next;
+        prev = block;
+        block = block->next;
+    }
+    if (!block) {
+        fprintf(stderr, "Bad ram offset %" PRIx64 "\n", (uint64_t)addr);
+        abort();
+    }
+    /* Move this entry to to start of the list.  */
+    if (prev) {
+        prev->next = block->next;
+        block->next = *prevp;
+        *prevp = block;
+    }
+    return block->host + (addr - block->offset);
 }
 
 /* Some of the softmmu routines need to translate from a host pointer
    (typically a TLB entry) back to a ram offset.  */
 ram_addr_t qemu_ram_addr_from_host(void *ptr)
 {
-  return (uint8_t *)ptr - phys_ram_base;
+    RAMBlock *prev;
+    RAMBlock **prevp;
+    RAMBlock *block;
+    uint8_t *host = ptr;
+
+#ifdef USE_KQEMU
+    if (kqemu_phys_ram_base) {
+        return host - kqemu_phys_ram_base;
+    }
+#endif
+
+    prev = NULL;
+    prevp = &ram_blocks;
+    block = ram_blocks;
+    while (block && (block->host > host
+                     || block->host + block->length <= host)) {
+        if (prev)
+          prevp = &prev->next;
+        prev = block;
+        block = block->next;
+    }
+    if (!block) {
+        fprintf(stderr, "Bad ram pointer %p\n", ptr);
+        abort();
+    }
+    return block->offset + (host - block->host);
 }
 
 static uint32_t unassigned_mem_readb(void *opaque, target_phys_addr_t addr)
@@ -2895,9 +2993,13 @@
 
     io_mem_watch = cpu_register_io_memory(0, watch_mem_read,
                                           watch_mem_write, NULL);
-    /* alloc dirty bits array */
-    phys_ram_dirty = qemu_vmalloc(phys_ram_size >> TARGET_PAGE_BITS);
-    memset(phys_ram_dirty, 0xff, phys_ram_size >> TARGET_PAGE_BITS);
+#ifdef USE_KQEMU
+    if (kqemu_phys_ram_base) {
+        /* alloc dirty bits array */
+        phys_ram_dirty = qemu_vmalloc(kqemu_phys_ram_size >> TARGET_PAGE_BITS);
+        memset(phys_ram_dirty, 0xff, kqemu_phys_ram_size >> TARGET_PAGE_BITS);
+    }
+#endif
 }
 
 /* mem_read and mem_write are arrays of functions containing the

Modified: trunk/kqemu.c
===================================================================
--- trunk/kqemu.c       2009-04-11 17:01:42 UTC (rev 7087)
+++ trunk/kqemu.c       2009-04-11 17:15:54 UTC (rev 7088)
@@ -91,6 +91,8 @@
 uint8_t *modified_ram_pages_table;
 int qpi_io_memory;
 uint32_t kqemu_comm_base; /* physical address of the QPI communication page */
+ram_addr_t kqemu_phys_ram_size;
+uint8_t *kqemu_phys_ram_base;
 
 #define cpuid(index, eax, ebx, ecx, edx) \
   asm volatile ("cpuid" \
@@ -214,13 +216,14 @@
                                       sizeof(uint64_t));
     if (!modified_ram_pages)
         goto fail;
-    modified_ram_pages_table = qemu_mallocz(phys_ram_size >> TARGET_PAGE_BITS);
+    modified_ram_pages_table =
+        qemu_mallocz(kqemu_phys_ram_size >> TARGET_PAGE_BITS);
     if (!modified_ram_pages_table)
         goto fail;
 
     memset(&kinit, 0, sizeof(kinit)); /* set the paddings to zero */
-    kinit.ram_base = phys_ram_base;
-    kinit.ram_size = phys_ram_size;
+    kinit.ram_base = kqemu_phys_ram_base;
+    kinit.ram_size = kqemu_phys_ram_size;
     kinit.ram_dirty = phys_ram_dirty;
     kinit.pages_to_flush = pages_to_flush;
     kinit.ram_pages_to_update = ram_pages_to_update;

Modified: trunk/vl.c
===================================================================
--- trunk/vl.c  2009-04-11 17:01:42 UTC (rev 7087)
+++ trunk/vl.c  2009-04-11 17:15:54 UTC (rev 7088)
@@ -3094,9 +3094,9 @@
     int ret;
     ram_addr_t i;
 
-    if (qemu_get_be32(f) != phys_ram_size)
+    if (qemu_get_be32(f) != last_ram_offset)
         return -EINVAL;
-    for(i = 0; i < phys_ram_size; i+= TARGET_PAGE_SIZE) {
+    for(i = 0; i < last_ram_offset; i+= TARGET_PAGE_SIZE) {
         ret = ram_get_page(f, qemu_get_ram_ptr(i), TARGET_PAGE_SIZE);
         if (ret)
             return ret;
@@ -3182,7 +3182,7 @@
     ram_addr_t addr = 0;
     int found = 0;
 
-    while (addr < phys_ram_size) {
+    while (addr < last_ram_offset) {
         if (cpu_physical_memory_get_dirty(current_addr, MIGRATION_DIRTY_FLAG)) 
{
             uint8_t *p;
 
@@ -3204,7 +3204,7 @@
             break;
         }
         addr += TARGET_PAGE_SIZE;
-        current_addr = (saved_addr + addr) % phys_ram_size;
+        current_addr = (saved_addr + addr) % last_ram_offset;
     }
 
     return found;
@@ -3217,7 +3217,7 @@
     ram_addr_t addr;
     ram_addr_t count = 0;
 
-    for (addr = 0; addr < phys_ram_size; addr += TARGET_PAGE_SIZE) {
+    for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
         if (cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
             count++;
     }
@@ -3231,7 +3231,7 @@
 
     if (stage == 1) {
         /* Make sure all dirty bits are set */
-        for (addr = 0; addr < phys_ram_size; addr += TARGET_PAGE_SIZE) {
+        for (addr = 0; addr < last_ram_offset; addr += TARGET_PAGE_SIZE) {
             if (!cpu_physical_memory_get_dirty(addr, MIGRATION_DIRTY_FLAG))
                 cpu_physical_memory_set_dirty(addr);
         }
@@ -3239,7 +3239,7 @@
         /* Enable dirty memory tracking */
         cpu_physical_memory_set_dirty_tracking(1);
 
-        qemu_put_be64(f, phys_ram_size | RAM_SAVE_FLAG_MEM_SIZE);
+        qemu_put_be64(f, last_ram_offset | RAM_SAVE_FLAG_MEM_SIZE);
     }
 
     while (!qemu_file_rate_limit(f)) {
@@ -3272,7 +3272,7 @@
 
     if (ram_decompress_open(s, f) < 0)
         return -EINVAL;
-    for(i = 0; i < phys_ram_size; i+= BDRV_HASH_BLOCK_SIZE) {
+    for(i = 0; i < last_ram_offset; i+= BDRV_HASH_BLOCK_SIZE) {
         if (ram_decompress_buf(s, buf, 1) < 0) {
             fprintf(stderr, "Error while reading ram block header\n");
             goto error;
@@ -3303,7 +3303,7 @@
         return ram_load_v1(f, opaque);
 
     if (version_id == 2) {
-        if (qemu_get_be32(f) != phys_ram_size)
+        if (qemu_get_be32(f) != last_ram_offset)
             return -EINVAL;
         return ram_load_dead(f, opaque);
     }
@@ -3318,7 +3318,7 @@
         addr &= TARGET_PAGE_MASK;
 
         if (flags & RAM_SAVE_FLAG_MEM_SIZE) {
-            if (addr != phys_ram_size)
+            if (addr != last_ram_offset)
                 return -EINVAL;
         }
 
@@ -5132,32 +5132,22 @@
             exit(1);
 
     /* init the memory */
-    phys_ram_size = machine->ram_require & ~RAMSIZE_FIXED;
+    if (ram_size == 0)
+        ram_size = DEFAULT_RAM_SIZE * 1024 * 1024;
 
-    if (machine->ram_require & RAMSIZE_FIXED) {
-        if (ram_size > 0) {
-            if (ram_size < phys_ram_size) {
-                fprintf(stderr, "Machine `%s' requires %llu bytes of memory\n",
-                                machine->name, (unsigned long long) 
phys_ram_size);
-                exit(-1);
-            }
-
-            phys_ram_size = ram_size;
-        } else
-            ram_size = phys_ram_size;
-    } else {
-        if (ram_size == 0)
-            ram_size = DEFAULT_RAM_SIZE * 1024 * 1024;
-
-        phys_ram_size += ram_size;
+#ifdef USE_KQEMU
+    /* FIXME: This is a nasty hack because kqemu can't cope with dynamic
+       guest ram allocation.  It needs to go away.  */
+    if (kqemu_allowed) {
+        kqemu_phys_ram_size = ram_size + VGA_RAM_SIZE + 4 * 1024 * 1024;
+        kqemu_phys_ram_base = qemu_vmalloc(kqemu_phys_ram_size);
+        if (!kqemu_phys_ram_base) {
+            fprintf(stderr, "Could not allocate physical memory\n");
+            exit(1);
+        }
     }
+#endif
 
-    phys_ram_base = qemu_vmalloc(phys_ram_size);
-    if (!phys_ram_base) {
-        fprintf(stderr, "Could not allocate physical memory\n");
-        exit(1);
-    }
-
     /* init the dynamic translator */
     cpu_exec_init_all(tb_size * 1024 * 1024);