[Qemu-devel] [PATCH 3/6] Convert multiboot to fw_cfg backed data storage

[Alexander Graf]

Right now we load the guest kernel to RAM, fire off the BIOS, hope it
doesn't clobber memory and run an option rom that jumps into the kernel.

That breaks with SeaBIOS, as that clears memory. So let's read all
kernel, module etc. data using the fw_cfg interface when in the int19
handler.

This patch implements said mechanism for multiboot.

Signed-off-by: Alexander Graf <address@hidden>

---

v1 -> v2:

  - rename copy_rom -> rom_copy
---
 hw/fw_cfg.h                   |    5 ++-
 hw/pc.c                       |   43 ++++++++++++++++-------
 pc-bios/optionrom/multiboot.S |   77 ++++++++++++++++++++++++++++++++---------
 3 files changed, 94 insertions(+), 31 deletions(-)

diff --git a/hw/fw_cfg.h b/hw/fw_cfg.h
index 359d45a..1e004b7 100644
--- a/hw/fw_cfg.h
+++ b/hw/fw_cfg.h
@@ -17,7 +17,10 @@
 #define FW_CFG_NUMA             0x0d
 #define FW_CFG_BOOT_MENU        0x0e
 #define FW_CFG_MAX_CPUS         0x0f
-#define FW_CFG_MAX_ENTRY        0x10
+#define FW_CFG_KERNEL_ENTRY     0x10
+#define FW_CFG_KERNEL_DATA      0x11
+#define FW_CFG_INITRD_DATA      0x12
+#define FW_CFG_MAX_ENTRY        0x13
 
 #define FW_CFG_WRITE_CHANNEL    0x4000
 #define FW_CFG_ARCH_LOCAL       0x8000
diff --git a/hw/pc.c b/hw/pc.c
index bf4718e..55bd1a4 100644
--- a/hw/pc.c
+++ b/hw/pc.c
@@ -603,6 +603,8 @@ static int load_multiboot(void *fw_cfg,
     uint32_t mb_mod_end;
     uint8_t bootinfo[0x500];
     uint32_t cmdline = 0x200;
+    uint8_t *mb_kernel_data;
+    uint8_t *mb_bootinfo_data;
 
     /* Ok, let's see if it is a multiboot image.
        The header is 12x32bit long, so the latest entry may be 8192 - 48. */
@@ -643,6 +645,12 @@ static int load_multiboot(void *fw_cfg,
         mh_load_addr = mh_entry_addr = elf_entry;
         mb_kernel_size = kernel_size;
 
+        mb_kernel_data = qemu_malloc(mb_kernel_size);
+        if (rom_copy(mb_kernel_data, elf_entry, kernel_size) != kernel_size) {
+            fprintf(stderr, "Error while fetching elf kernel from rom\n");
+            exit(1);
+        }
+
 #ifdef DEBUG_MULTIBOOT
         fprintf(stderr, "qemu: loading multiboot-elf kernel (%#x bytes) with 
entry %#zx\n",
                 mb_kernel_size, (size_t)mh_entry_addr);
@@ -656,7 +664,6 @@ static int load_multiboot(void *fw_cfg,
         uint32_t mh_bss_end_addr = ldl_p(header+i+24);
 #endif
         uint32_t mb_kernel_text_offset = i - (mh_header_addr - mh_load_addr);
-        uint8_t *kernel;
 
         mh_entry_addr = ldl_p(header+i+28);
         mb_kernel_size = get_file_size(f) - mb_kernel_text_offset;
@@ -676,12 +683,9 @@ static int load_multiboot(void *fw_cfg,
                 mb_kernel_size, mh_load_addr);
 #endif
 
-        kernel = qemu_malloc(mb_kernel_size);
+        mb_kernel_data = qemu_malloc(mb_kernel_size);
         fseek(f, mb_kernel_text_offset, SEEK_SET);
-        fread(kernel, 1, mb_kernel_size, f);
-        rom_add_blob_fixed(kernel_filename, kernel, mb_kernel_size,
-                           mh_load_addr);
-        qemu_free(kernel);
+        fread(mb_kernel_data, 1, mb_kernel_size, f);
         fclose(f);
     }
 
@@ -732,9 +736,14 @@ static int load_multiboot(void *fw_cfg,
                 exit(1);
             }
             mb_mod_end = mb_mod_start + mb_mod_length;
-            rom_add_file_fixed(initrd_filename, mb_mod_start);
-
             mb_mod_count++;
+
+            /* append module data at the end of last module */
+            mb_kernel_data = qemu_realloc(mb_kernel_data,
+                                          mh_load_addr - mb_mod_end);
+            load_image(initrd_filename,
+                       mb_kernel_data + mb_mod_start - mh_load_addr);
+
             stl_p(bootinfo + mb_mod_info + 0, mb_mod_start);
             stl_p(bootinfo + mb_mod_info + 4, mb_mod_start + mb_mod_length);
             stl_p(bootinfo + mb_mod_info + 12, 0x0); /* reserved */
@@ -774,13 +783,21 @@ static int load_multiboot(void *fw_cfg,
     fprintf(stderr, "multiboot: mh_entry_addr = %#x\n", mh_entry_addr);
 #endif
 
+    /* save bootinfo off the stack */
+    mb_bootinfo_data = qemu_malloc(sizeof(bootinfo));
+    memcpy(mb_bootinfo_data, bootinfo, sizeof(bootinfo));
+
     /* Pass variables to option rom */
-    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_entry_addr);
-    fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, mb_bootinfo);
-    fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, mmap_addr);
+    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ENTRY, mh_entry_addr);
+    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_ADDR, mh_load_addr);
+    fw_cfg_add_i32(fw_cfg, FW_CFG_KERNEL_SIZE, mb_mod_end - mh_load_addr);
+    fw_cfg_add_bytes(fw_cfg, FW_CFG_KERNEL_DATA, mb_kernel_data,
+                     mb_mod_end - mh_load_addr);
 
-    rom_add_blob_fixed("multiboot-info", bootinfo, sizeof(bootinfo),
-                       mb_bootinfo);
+    fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_ADDR, mb_bootinfo);
+    fw_cfg_add_i32(fw_cfg, FW_CFG_INITRD_SIZE, sizeof(bootinfo));
+    fw_cfg_add_bytes(fw_cfg, FW_CFG_INITRD_DATA, mb_bootinfo_data,
+                     sizeof(bootinfo));
 
     option_rom[nb_option_roms] = "multiboot.bin";
     nb_option_roms++;
diff --git a/pc-bios/optionrom/multiboot.S b/pc-bios/optionrom/multiboot.S
index e6cbefd..dafac73 100644
--- a/pc-bios/optionrom/multiboot.S
+++ b/pc-bios/optionrom/multiboot.S
@@ -26,6 +26,14 @@
 
 #define MULTIBOOT_MAGIC                0x2badb002
 
+#define GS_PROT_JUMP           0
+#define GS_GDT_DESC            6
+
+/* Break the translation block flow so -d cpu shows us values */
+#define DEBUG_HERE \
+       jmp             1f;                             \
+       1:
+       
 /* Read a variable from the fw_cfg device.
    Clobbers:   %edx
    Out:                %eax */
@@ -44,12 +52,31 @@
        bswap           %eax
 .endm
 
+/*
+ * Read a blob from the fw_cfg device.
+ * Requires _ADDR, _SIZE and _DATA values for the parameter.
+ *
+ * Clobbers:   %eax, %edx, %es, %ecx, %edi
+ */
+#define read_fw_blob(var) \
+       read_fw         var ## _ADDR;                   \
+       mov             %eax, %edi;                     \
+       read_fw         var ## _SIZE;                   \
+       mov             %eax, %ecx;                     \
+       mov             $var ## _DATA, %ax;             \
+       mov             $BIOS_CFG_IOPORT_CFG, %edx;     \
+       outw            %ax, (%dx);                     \
+       mov             $BIOS_CFG_IOPORT_DATA, %dx;     \
+       cld;                                            \
+       DEBUG_HERE \
+       rep insb        (%dx), %es:(%edi);
+
 .code16
 .text
        .global         _start
 _start:
        .short          0xaa55
-       .byte           1 /* (_end - _start) / 512 */
+       .byte           (_end - _start) / 512
        push            %eax
        push            %ds
 
@@ -57,10 +84,6 @@ _start:
        xor             %ax, %ax
        mov             %ax, %ds
 
-       /* save old int 19 */
-       mov             (0x19*4), %eax
-       mov             %eax, %cs:old_int19
-
        /* install our int 19 handler */
        movw            $int19_handler, (0x19*4)
        mov             %cs, (0x19*4+2)
@@ -84,15 +107,34 @@ run_multiboot:
        mov             %cs, %eax
        shl             $0x4, %eax
 
-       /* fix the gdt descriptor to be PC relative */
-       mov             (gdt_desc+2), %ebx
-       add             %eax, %ebx
-       mov             %ebx, (gdt_desc+2)
+       /* set up a long jump descriptor that is PC relative */
 
-       /* fix the prot mode indirect jump to be PC relative */
+       /* move stack memory to %gs */
+       mov             %ss, %ecx
+       shl             $0x4, %ecx
+       mov             %esp, %ebx
+       add             %ebx, %ecx
+       sub             $0x20, %ecx
+       sub             $0x30, %esp
+       shr             $0x4, %ecx
+       mov             %cx, %gs
+
+       /* now push the indirect jump decriptor there */
        mov             (prot_jump), %ebx
        add             %eax, %ebx
-       mov             %ebx, (prot_jump)
+       movl            %ebx, %gs:GS_PROT_JUMP
+       mov             $8, %bx
+       movw            %bx, %gs:GS_PROT_JUMP + 4
+
+       /* fix the gdt descriptor to be PC relative */
+       movw            (gdt_desc), %bx
+       movw            %bx, %gs:GS_GDT_DESC
+       movl            (gdt_desc+2), %ebx
+       add             %eax, %ebx
+       movl            %ebx, %gs:GS_GDT_DESC + 2
+
+       /* Read the bootinfo struct into RAM */
+       read_fw_blob(FW_CFG_INITRD)
 
        /* FS = bootinfo_struct */
        read_fw         FW_CFG_INITRD_ADDR
@@ -100,7 +142,7 @@ run_multiboot:
        mov             %ax, %fs
 
        /* ES = mmap_addr */
-       read_fw         FW_CFG_INITRD_SIZE
+       mov             %eax, %fs:0x48
        shr             $4, %eax
        mov             %ax, %es
 
@@ -144,7 +186,7 @@ mmap_done:
 real_to_prot:
        /* Load the GDT before going into protected mode */
 lgdt:
-       data32 lgdt     %cs:gdt_desc
+       data32 lgdt     %gs:GS_GDT_DESC
 
        /* get us to protected mode now */
        movl            $1, %eax
@@ -152,7 +194,7 @@ lgdt:
 
        /* the LJMP sets CS for us and gets us to 32-bit */
 ljmp:
-       data32 ljmp     *%cs:prot_jump
+       data32 ljmp     *%gs:GS_PROT_JUMP
 
 prot_mode:
 .code32
@@ -165,8 +207,11 @@ prot_mode:
        movl            %eax, %fs
        movl            %eax, %gs
 
+       /* Read the kernel and modules into RAM */
+       read_fw_blob(FW_CFG_KERNEL)
+
        /* Jump off to the kernel */
-       read_fw         FW_CFG_KERNEL_ADDR
+       read_fw         FW_CFG_KERNEL_ENTRY
        mov             %eax, %ecx
 
        /* EBX contains a pointer to the bootinfo struct */
@@ -180,8 +225,6 @@ ljmp2:
 
 /* Variables */
 .align 4, 0
-old_int19:     .long 0
-
 prot_jump:     .long prot_mode
                .short 8
 
-- 
1.6.0.2