Reading Linux from the fw_cfg interface is faster using the DMA interface. For this reason, add a Linux loader that can benefit from this interface.
Signed-off-by: Marc Marí markmb@redhat.com --- src/boot.c | 26 +++++++++++++++++ src/fw/paravirt.c | 84 ++++++++++++++++++++++++++++++++++++++++++++----------- src/fw/paravirt.h | 35 +++++++++++++++++++++++ src/romlayout.S | 20 +++++++++++++ src/util.h | 1 + 5 files changed, 150 insertions(+), 16 deletions(-)
diff --git a/src/boot.c b/src/boot.c index e0f73a3..ba692db 100644 --- a/src/boot.c +++ b/src/boot.c @@ -280,6 +280,14 @@ boot_init(void) BootRetryTime = romfile_loadint("etc/boot-fail-wait", 60*1000);
loadBootOrder(); + + /* Check for booting directly from fw_cfg DMA. We assign the same boot + * priority of the linuxboot rom (if it exists). + */ + if (qemu_cfg_dma_enabled()) { + boot_add_dma("QEMU Kernel image", + bootprio_find_named_rom("genroms/linuxboot.bin", 0)); + } }
@@ -304,6 +312,7 @@ static struct hlist_head BootList VARVERIFY32INIT; #define IPL_TYPE_HARDDISK 0x02 #define IPL_TYPE_CDROM 0x03 #define IPL_TYPE_CBFS 0x20 +#define IPL_TYPE_DMA 0x21 #define IPL_TYPE_BEV 0x80 #define IPL_TYPE_BCV 0x81 #define IPL_TYPE_HALT 0xf0 @@ -398,6 +407,12 @@ boot_add_cbfs(void *data, const char *desc, int prio) bootentry_add(IPL_TYPE_CBFS, defPrio(prio, DEFAULT_PRIO), (u32)data, desc); }
+// Add a FW_CFG DMA boot +void +boot_add_dma(const char *desc, int prio) +{ + bootentry_add(IPL_TYPE_DMA, defPrio(prio, DEFAULT_PRIO), 0, desc); +}
/**************************************************************** * Keyboard calls @@ -684,6 +699,14 @@ boot_rom(u32 vector) call_boot_entry(so, 0); }
+// Boot from a linuxboot ROM when QEMU cfg is in DMA mode +static void +boot_linux_cfg_dma(void) +{ + printf("Booting Linux from fw_cfg...\n"); + qemu_cfg_dma_boot_linux(); +} + // Unable to find bootable device - warn user and eventually retry. static void boot_fail(void) @@ -734,6 +757,9 @@ do_boot(int seq_nr) case IPL_TYPE_BEV: boot_rom(ie->vector); break; + case IPL_TYPE_DMA: + boot_linux_cfg_dma(); + break; case IPL_TYPE_HALT: boot_fail(); break; diff --git a/src/fw/paravirt.c b/src/fw/paravirt.c index 26af499..797d8ba 100644 --- a/src/fw/paravirt.c +++ b/src/fw/paravirt.c @@ -23,7 +23,8 @@ #include "util.h" // pci_setup #include "x86.h" // cpuid #include "xen.h" // xen_biostable_setup -#include "stacks.h" // yield +#include "bregs.h" // struct bregs +#include "stacks.h" // farcall16big, yield
// Amount of continuous ram under 4Gig u32 RamSize; @@ -185,21 +186,6 @@ qemu_platform_setup(void) * QEMU firmware config (fw_cfg) interface ****************************************************************/
-// List of QEMU fw_cfg entries. DO NOT ADD MORE. (All new content -// should be passed via the fw_cfg "file" interface.) -#define QEMU_CFG_SIGNATURE 0x00 -#define QEMU_CFG_ID 0x01 -#define QEMU_CFG_UUID 0x02 -#define QEMU_CFG_NUMA 0x0d -#define QEMU_CFG_BOOT_MENU 0x0e -#define QEMU_CFG_MAX_CPUS 0x0f -#define QEMU_CFG_FILE_DIR 0x19 -#define QEMU_CFG_ARCH_LOCAL 0x8000 -#define QEMU_CFG_ACPI_TABLES (QEMU_CFG_ARCH_LOCAL + 0) -#define QEMU_CFG_SMBIOS_ENTRIES (QEMU_CFG_ARCH_LOCAL + 1) -#define QEMU_CFG_IRQ0_OVERRIDE (QEMU_CFG_ARCH_LOCAL + 2) -#define QEMU_CFG_E820_TABLE (QEMU_CFG_ARCH_LOCAL + 3) - static void qemu_cfg_select(u16 f) { @@ -505,3 +491,67 @@ void qemu_cfg_init(void) dprintf(1, "Moving pm_base to 0x%x\n", acpi_pm_base); } } + +void qemu_cfg_dma_boot_linux(void) +{ + dprintf(1, "Loading kernel\n"); + void *setup_addr; + u32 setup_size; + qemu_cfg_read_entry(&setup_addr, QEMU_CFG_SETUP_ADDR, 4); + qemu_cfg_read_entry(&setup_size, QEMU_CFG_SETUP_SIZE, 4); + qemu_cfg_read_entry(setup_addr, QEMU_CFG_SETUP_DATA, setup_size); + + if (readl(setup_addr + 0x202) != 0x53726448) { + dprintf(1, "Not valid kernel\n"); + return; + } + + u16 protocol = readw(setup_addr + 0x206); + if (protocol < 0x203) { + dprintf(1, "Old kernel (v %x) not supported\n", protocol); + return; + } + + void *kernel_addr; + u32 kernel_size; + + qemu_cfg_read_entry(&kernel_addr, QEMU_CFG_KERNEL_ADDR, 4); + qemu_cfg_read_entry(&kernel_size, QEMU_CFG_KERNEL_SIZE, 4); + qemu_cfg_read_entry(kernel_addr, QEMU_CFG_KERNEL_DATA, kernel_size); + + void *cmdline_addr; + u32 cmdline_size; + qemu_cfg_read_entry(&cmdline_addr, QEMU_CFG_CMDLINE_ADDR, 4); + qemu_cfg_read_entry(&cmdline_size, QEMU_CFG_CMDLINE_SIZE, 4); + if (cmdline_size) { + qemu_cfg_read_entry(cmdline_addr, QEMU_CFG_CMDLINE_DATA, cmdline_size); + } + + void *initrd_addr; + u32 initrd_size; + qemu_cfg_read_entry(&initrd_addr, QEMU_CFG_INITRD_ADDR, 4); + qemu_cfg_read_entry(&initrd_size, QEMU_CFG_INITRD_SIZE, 4); + if (initrd_size) { + qemu_cfg_read_entry(initrd_addr, QEMU_CFG_INITRD_DATA, initrd_size); + } + + // Last configurations + writel(setup_addr + 0x228, (u32)cmdline_addr); + writeb(setup_addr + 0x210, 0xB0); + writeb(setup_addr + 0x211, readb(setup_addr + 0x211) | 0x80); + writew(setup_addr + 0x224, cmdline_addr - setup_addr - 0x200); + writel(setup_addr + 0x218, (u32)initrd_addr); + writel(setup_addr + 0x21c, initrd_size); + + dprintf(1, "Jumping to kernel %d@%x %d@%x %d@%x %d@%x\n" + , setup_size, (u32)setup_addr, cmdline_size, (u32)cmdline_addr + , kernel_size, (u32)kernel_addr, initrd_size, (u32)initrd_addr); + struct bregs br; + memset(&br, 0, sizeof(br)); + extern void kernel_stub(void); + br.ebx = (u32)setup_addr >> 4; + br.edx = (u32)cmdline_addr - (u32)setup_addr - 16; + br.code = SEGOFF(SEG_BIOS, (u32)kernel_stub - BUILD_BIOS_ADDR); + + farcall16big(&br); +} diff --git a/src/fw/paravirt.h b/src/fw/paravirt.h index 5e5c61e..a9674c0 100644 --- a/src/fw/paravirt.h +++ b/src/fw/paravirt.h @@ -46,9 +46,44 @@ static inline int runningOnKVM(void) { // QEMU_CFG_DMA ID bit #define QEMU_CFG_VERSION_DMA 2
+// List of QEMU fw_cfg entries. DO NOT ADD MORE. (All new content +// should be passed via the fw_cfg "file" interface.) +#define QEMU_CFG_SIGNATURE 0x00 +#define QEMU_CFG_ID 0x01 +#define QEMU_CFG_UUID 0x02 +#define QEMU_CFG_RAM_SIZE 0x03 +#define QEMU_CFG_NOGRAPHIC 0x04 +#define QEMU_CFG_NB_CPUS 0x05 +#define QEMU_CFG_MACHINE_ID 0x06 +#define QEMU_CFG_KERNEL_ADDR 0x07 +#define QEMU_CFG_KERNEL_SIZE 0x08 +#define QEMU_CFG_KERNEL_CMDLINE 0x09 +#define QEMU_CFG_INITRD_ADDR 0x0a +#define QEMU_CFG_INITRD_SIZE 0x0b +#define QEMU_CFG_BOOT_DEVICE 0x0c +#define QEMU_CFG_NUMA 0x0d +#define QEMU_CFG_BOOT_MENU 0x0e +#define QEMU_CFG_MAX_CPUS 0x0f +#define QEMU_CFG_KERNEL_ENTRY 0x10 +#define QEMU_CFG_KERNEL_DATA 0x11 +#define QEMU_CFG_INITRD_DATA 0x12 +#define QEMU_CFG_CMDLINE_ADDR 0x13 +#define QEMU_CFG_CMDLINE_SIZE 0x14 +#define QEMU_CFG_CMDLINE_DATA 0x15 +#define QEMU_CFG_SETUP_ADDR 0x16 +#define QEMU_CFG_SETUP_SIZE 0x17 +#define QEMU_CFG_SETUP_DATA 0x18 +#define QEMU_CFG_FILE_DIR 0x19 +#define QEMU_CFG_ARCH_LOCAL 0x8000 +#define QEMU_CFG_ACPI_TABLES (QEMU_CFG_ARCH_LOCAL + 0) +#define QEMU_CFG_SMBIOS_ENTRIES (QEMU_CFG_ARCH_LOCAL + 1) +#define QEMU_CFG_IRQ0_OVERRIDE (QEMU_CFG_ARCH_LOCAL + 2) +#define QEMU_CFG_E820_TABLE (QEMU_CFG_ARCH_LOCAL + 3) + int qemu_cfg_dma_enabled(void); void qemu_preinit(void); void qemu_platform_setup(void); void qemu_cfg_init(void); +void qemu_cfg_dma_boot_linux(void);
#endif diff --git a/src/romlayout.S b/src/romlayout.S index 7938e22..1c641c2 100644 --- a/src/romlayout.S +++ b/src/romlayout.S @@ -196,6 +196,26 @@ __farcall16: IRQ_TRAMPOLINE 1c IRQ_TRAMPOLINE 4a
+ DECLFUNC kernel_stub +kernel_stub: + movw %bx, %ds + movw %bx, %es + movw %bx, %fs + movw %bx, %gs + movw %bx, %ss + movl %edx, %esp + addw $0x20, %bx + pushw %bx // push CS + pushw $0 // push IP + xorl %eax, %eax + xorl %ebx, %ebx + xorl %ecx, %ecx + xorl %edx, %edx + xorl %edi, %edi + xorl %esi, %esi + xorl %ebp, %ebp + lretw +
/**************************************************************** * Misc. entry points. diff --git a/src/util.h b/src/util.h index 327abeb..f99cdb5 100644 --- a/src/util.h +++ b/src/util.h @@ -25,6 +25,7 @@ void boot_add_floppy(struct drive_s *drive_g, const char *desc, int prio); void boot_add_hd(struct drive_s *drive_g, const char *desc, int prio); void boot_add_cd(struct drive_s *drive_g, const char *desc, int prio); void boot_add_cbfs(void *data, const char *desc, int prio); +void boot_add_dma(const char *desc, int prio); void interactive_bootmenu(void); void bcv_prepboot(void); struct pci_device;