[Qemu-devel] [PATCH 16/40] xenner: kernel: Main (i386)

[Alexander Graf]

This patch adds the i386 specific piece of xenner's main loop.

Signed-off-by: Alexander Graf <address@hidden>
---
 pc-bios/xenner/xenner-main32.c |  390 ++++++++++++++++++++++++++++++++++++++++
 1 files changed, 390 insertions(+), 0 deletions(-)
 create mode 100644 pc-bios/xenner/xenner-main32.c

diff --git a/pc-bios/xenner/xenner-main32.c b/pc-bios/xenner/xenner-main32.c
new file mode 100644
index 0000000..0c049dd
--- /dev/null
+++ b/pc-bios/xenner/xenner-main32.c
@@ -0,0 +1,390 @@
+/*
+ *  Copyright (C) Red Hat 2007
+ *  Copyright (C) Novell Inc. 2010
+ *
+ *  Author(s): Gerd Hoffmann <address@hidden>
+ *             Alexander Graf <address@hidden>
+ *
+ *  Xenner main functions for 32 bit
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; under version 2 of the License.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License along
+ *  with this program; if not, see <http://www.gnu.org/licenses/>.
+ */
+
+#include "xenner.h"
+#include "xenner-main.c"
+
+/* --------------------------------------------------------------------- */
+/* helper functions                                                      */
+
+int bounce_trap(struct xen_cpu *cpu, struct regs_32 *regs, int trapno, int 
cbno)
+{
+    uint32_t *kesp, eip = 0, cs = 0;
+    uint32_t stack_cs, stack_eflags;
+    int stack_switch = 0;
+    int error_code = 0;
+    int interrupt = 0;
+    int k = 0;
+
+    vminfo.faults[XEN_FAULT_BOUNCE_TRAP]++;
+
+    if (trapno >= 0) {
+        /* trap bounce */
+        eip  = xentr[trapno].address;
+        cs   = xentr[trapno].cs;
+        if (TI_GET_IF(&xentr[trapno])) {
+            interrupt = 1;
+        }
+        if (trapno < (sizeof(trapinfo) / sizeof(trapinfo[0]))) {
+            error_code = trapinfo[trapno].ec;
+        }
+        if (trapno == 14) {
+            /* page fault */
+            cpu->v.vcpu_info->arch.cr2 = read_cr2();
+        }
+    }
+    if (cbno >= 0) {
+        /* callback */
+        eip  = xencb[cbno].eip;
+        cs   = xencb[cbno].cs;
+        switch (cbno) {
+        case CALLBACKTYPE_event:
+            interrupt = 1;
+            break;
+        }
+    }
+
+    if (!cs) {
+        printk(0, "%s: trapno %d, cbno %d\n", __FUNCTION__, trapno, cbno);
+        panic("no guest trap handler", regs);
+    }
+
+    /* set interrupt flag depending on event channel mask */
+    stack_eflags = regs->eflags & ~X86_EFLAGS_IF;
+    if (guest_irq_flag(cpu)) {
+        stack_eflags |= X86_EFLAGS_IF;
+    }
+
+    /* old evtchn_upcall_mask is saved in cs slot on the stack */
+    stack_cs = regs->cs | ((uint32_t)cpu->v.vcpu_info->evtchn_upcall_mask << 
16);
+    if (interrupt) {
+        guest_cli(cpu);
+    }
+
+    if ((regs->cs & 0x03) < (cs & 0x03)) {
+        panic("bounce trap: illegal ring switch", regs);
+    }
+    if ((regs->cs & 0x03) > (cs & 0x03)) {
+        stack_switch = 1;
+    }
+
+    /* prepare guest stack: copy from emu, so the handler
+     * jumps straigt back without round-trip via emu */
+    if (stack_switch) {
+        kesp = (void*)cpu->tss.esp1;
+        kesp[-(++k)] = regs->ss;            // push ss
+        kesp[-(++k)] = regs->esp;           // push esp
+    } else {
+        kesp = (void*)regs->esp;
+    }
+
+    kesp[-(++k)] = stack_eflags;            // push eflags
+    kesp[-(++k)] = stack_cs;                // push cs
+    kesp[-(++k)] = regs->eip;               // push eip
+    if (error_code) {
+        kesp[-(++k)] = regs->error;         // push error code
+    }
+
+    /* prepare emu stack, so iret jumps to the kernels handler. */
+    regs->eip     = eip;
+    regs->cs      = cs;
+    regs->eflags &= EFLAGS_TRAPMASK;
+    if (stack_switch) {
+        regs->ss  = cpu->tss.ss1;
+        regs->esp = cpu->tss.esp1;
+    }
+    regs->esp -= 4*k;
+
+    return 0;
+}
+
+/* --------------------------------------------------------------------- */
+
+static const struct kvm_segment xen32_cs0 = {
+    .base     = 0,
+    .limit    = 0xffffffff,
+    .selector = 0xe008,
+    .dpl      = 0,
+    .type     = 0xb,
+    .present  = 1,  .db = 1,  .s = 1,  .g = 1,
+};
+static const struct kvm_segment xen32_ds0 = {
+    .base     = 0,
+    .limit    = 0xffffffff,
+    .selector = 0xe010,
+    .dpl      = 0,
+    .type     = 0x3,
+    .present  = 1,  .db = 1,  .s = 1,  .g = 1,
+};
+static const struct kvm_segment xen32_cs1 = {
+    .base     = 0,
+    .limit    = 0xffffffff,
+    .selector = 0xe019,
+    .dpl      = 1,
+    .type     = 0xb,
+    .present  = 1,  .db = 1,  .s = 1,  .g = 1,
+};
+static const struct kvm_segment xen32_ds1 = {
+    .base     = 0,
+    .limit    = 0xffffffff,
+    .selector = 0xe021,
+    .dpl      = 1,
+    .type     = 0x3,
+    .present  = 1,  .db = 1,  .s = 1,  .g = 1,
+};
+static const struct kvm_segment xen32_cs3 = {
+    .base     = 0,
+    .limit    = 0xffffffff,
+    .selector = 0xe02b,
+    .dpl      = 3,
+    .type     = 0xb,
+    .present  = 1,  .db = 1,  .s = 1,  .g = 1,
+};
+static const struct kvm_segment xen32_ds3 = {
+    .base     = 0,
+    .limit    = 0xffffffff,
+    .selector = 0xe033,
+    .dpl      = 3,
+    .type     = 0x3,
+    .present  = 1,  .db = 1,  .s = 1,  .g = 1,
+};
+
+void gdt_init(struct xen_cpu *cpu)
+{
+    printk(2, "%s: cpu %d\n", __FUNCTION__, cpu->id);
+
+    if (!cpu->gdt) {
+        cpu->gdt = get_pages(16, "gdt");
+    }
+
+    gdt_set(cpu->gdt, &xen32_cs0);
+    gdt_set(cpu->gdt, &xen32_ds0);
+    gdt_set(cpu->gdt, &xen32_cs1);
+    gdt_set(cpu->gdt, &xen32_ds1);
+    gdt_set(cpu->gdt, &xen32_cs3);
+    gdt_set(cpu->gdt, &xen32_ds3);
+}
+
+void tss_init(struct xen_cpu *cpu)
+{
+    struct descriptor_32 *gdt = cpu->gdt;
+    int idx = tss(cpu);
+
+    printk(2, "%s: cpu %d\n", __FUNCTION__, cpu->id);
+
+    cpu->tss.esp0 = (uintptr_t)cpu->stack_high;
+    cpu->tss.ss0  = 0xe010;
+
+    gdt[ idx ] = mkdesc32((uintptr_t)(&cpu->tss), sizeof(cpu->tss)-1, 0x89, 0);
+}
+
+void msrs_init(struct xen_cpu *cpu)
+{
+    printk(2, "%s: cpu %d\n", __FUNCTION__, cpu->id);
+}
+
+void idt_init(void)
+{
+    intptr_t entry;
+    int i,len;
+
+    printk(2, "%s\n", __FUNCTION__);
+
+    len = (irq_common - irq_entries) / 256;
+    for (i = 0; i < 256; i++) {
+        entry = (intptr_t)(irq_entries + i*len);
+        xen_idt[i] = mkgate32(0xe008, (uintptr_t)irq_entries + i*len, 0x8e);
+    }
+
+    xen_idt[    0 ] = mkgate32(0xe008, (uintptr_t)division_by_zero,    0x8e);
+    xen_idt[    1 ] = mkgate32(0xe008, (uintptr_t)debug_int1,          0x8e);
+    xen_idt[    2 ] = mkgate32(0xe008, (uintptr_t)nmi,                 0x8e);
+    xen_idt[    3 ] = mkgate32(0xe008, (uintptr_t)debug_int3,          0xee);
+    xen_idt[    4 ] = mkgate32(0xe008, (uintptr_t)overflow,            0x8e);
+    xen_idt[    5 ] = mkgate32(0xe008, (uintptr_t)bound_check,         0x8e);
+    xen_idt[    6 ] = mkgate32(0xe008, (uintptr_t)illegal_instruction, 0x8e);
+    xen_idt[    7 ] = mkgate32(0xe008, (uintptr_t)no_device,           0x8e);
+    xen_idt[    8 ] = mkgate32(0xe008, (uintptr_t)double_fault,        0x8e);
+    xen_idt[    9 ] = mkgate32(0xe008, (uintptr_t)coprocessor,         0x8e);
+    xen_idt[   10 ] = mkgate32(0xe008, (uintptr_t)invalid_tss,         0x8e);
+    xen_idt[   11 ] = mkgate32(0xe008, (uintptr_t)segment_not_present, 0x8e);
+    xen_idt[   12 ] = mkgate32(0xe008, (uintptr_t)stack_fault,         0x8e);
+    xen_idt[   13 ] = mkgate32(0xe008, (uintptr_t)general_protection,  0x8e);
+    xen_idt[   14 ] = mkgate32(0xe008, (uintptr_t)page_fault,          0x8e);
+    xen_idt[   16 ] = mkgate32(0xe008, (uintptr_t)floating_point,      0x8e);
+    xen_idt[   17 ] = mkgate32(0xe008, (uintptr_t)alignment,           0x8e);
+    xen_idt[   18 ] = mkgate32(0xe008, (uintptr_t)machine_check,       0x8e);
+    xen_idt[   19 ] = mkgate32(0xe008, (uintptr_t)simd_floating_point, 0x8e);
+
+    xen_idt[ VECTOR_FLUSH_TLB  ] =
+        mkgate32(0xe008, (uintptr_t)smp_flush_tlb, 0x8e);
+
+    xen_idt[ 0x82 ] = mkgate32(0xe008, (uintptr_t)xen_hypercall,       0xae);
+}
+
+/* --------------------------------------------------------------------- */
+
+static int pf_fixup_readonly(struct regs_32 *regs, uint32_t cr2)
+{
+    pte_t *pte = find_pte_lpt(cr2);
+
+    if (cr2 >= XEN_IPT) {
+        return 0;
+    }
+
+    if (*pte & _PAGE_USER) {
+        return 0;
+    }
+
+    /* is kernel page */
+    *pte |= _PAGE_RW;
+
+    vminfo.faults[XEN_FAULT_PAGE_FAULT_FIX_RO]++;
+    flush_tlb_addr(cr2);
+
+    return 1;
+}
+
+void guest_regs_init(struct xen_cpu *cpu, struct regs_32 *regs)
+{
+    struct vcpu_guest_context *ctxt = cpu->init_ctxt;
+
+    regs->eax    = ctxt->user_regs.eax;
+    regs->ebx    = ctxt->user_regs.ebx;
+    regs->ecx    = ctxt->user_regs.ecx;
+    regs->edx    = ctxt->user_regs.edx;
+    regs->esi    = ctxt->user_regs.esi;
+    regs->edi    = ctxt->user_regs.edi;
+    regs->ebp    = ctxt->user_regs.ebp;
+    regs->eip    = ctxt->user_regs.eip;
+    regs->cs     = ctxt->user_regs.cs;
+    regs->eflags = ctxt->user_regs.eflags;
+    regs->esp    = ctxt->user_regs.esp;
+    regs->ss     = ctxt->user_regs.ss;
+
+    regs->ds     = ctxt->user_regs.ds;
+    regs->es     = ctxt->user_regs.es;
+    asm volatile("mov %0, %%fs;\n" :: "r" (ctxt->user_regs.fs) : "memory");
+    asm volatile("mov %0, %%gs;\n" :: "r" (ctxt->user_regs.gs) : "memory");
+}
+
+static void set_up_shared_info(void)
+{
+    int i;
+
+    memset(&shared_info, 0, sizeof(shared_info));
+    for ( i = 0; i < XEN_LEGACY_MAX_VCPUS; i++ ) {
+        shared_info.vcpu_info[i].evtchn_upcall_mask = 1;
+    }
+}
+
+static void set_up_context(void *_ctxt, unsigned long boot_cr3,
+                           unsigned long init_pt_len)
+{
+    vcpu_guest_context_t *ctxt = _ctxt;
+    uint64_t virt_base = emudev_get(EMUDEV_CONF_PV_VIRT_BASE, 0);
+    uint64_t virt_entry = emudev_get(EMUDEV_CONF_PV_VIRT_ENTRY, 0);
+    uint64_t boot_stack_pfn = emudev_get(EMUDEV_CONF_PFN_INIT_PT, 0) +
+                              addr_to_frame(init_pt_len + PAGE_SIZE - 1);
+    uint64_t start_info_pfn = emudev_get(EMUDEV_CONF_PFN_START_INFO, 0);
+
+    set_up_shared_info();
+
+    /* clear everything */
+    memset(ctxt, 0, sizeof(*ctxt));
+
+    ctxt->user_regs.ds = FLAT_KERNEL_DS_X86_32;
+    ctxt->user_regs.es = FLAT_KERNEL_DS_X86_32;
+    ctxt->user_regs.fs = FLAT_KERNEL_DS_X86_32;
+    ctxt->user_regs.gs = FLAT_KERNEL_DS_X86_32;
+    ctxt->user_regs.ss = FLAT_KERNEL_SS_X86_32;
+    ctxt->user_regs.cs = FLAT_KERNEL_CS_X86_32;
+    ctxt->user_regs.eip = virt_entry;
+    ctxt->user_regs.esp = virt_base | ((boot_stack_pfn + 1) << PAGE_SHIFT);
+    ctxt->user_regs.esi = virt_base | (start_info_pfn << PAGE_SHIFT);
+    ctxt->user_regs.eflags = 1 << 9; /* Interrupt Enable */
+
+    ctxt->kernel_ss = ctxt->user_regs.ss;
+    ctxt->kernel_sp = ctxt->user_regs.esp;
+
+    ctxt->flags = VGCF_in_kernel_X86_32 | VGCF_online_X86_32;
+    ctxt->ctrlreg[3] = boot_cr3;
+}
+
+static void guest_hypercall_page(struct xen_cpu *cpu)
+{
+    unsigned long _hypercall_page = emudev_get(EMUDEV_CONF_HYPERCALL_PAGE, 0);
+    char *hypercall_page = (char*)_hypercall_page;
+
+    char *p;
+    int i;
+
+    /* Fill in all the transfer points with template machine code. */
+    for ( i = 0; i < (PAGE_SIZE / 32); i++ ) {
+        p = (char *)(hypercall_page + (i * 32));
+        *(uint8_t  *)(p+ 0) = 0xb8;    /* mov  $<i>,%eax */
+        *(uint32_t *)(p+ 1) = i;
+        *(uint16_t *)(p+ 5) = 0x82cd;  /* int  $0x82 */
+        *(uint8_t  *)(p+ 7) = 0xc3;    /* ret */
+    }
+
+    /*
+     * HYPERVISOR_iret is special because it doesn't return and expects a
+     * special stack frame. Guests jump at this transfer point instead of
+     * calling it.
+     */
+    p = (char *)(hypercall_page + (__HYPERVISOR_iret * 32));
+    *(uint8_t  *)(p+ 0) = 0x50;    /* push %eax */
+    *(uint8_t  *)(p+ 1) = 0xb8;    /* mov  $__HYPERVISOR_iret,%eax */
+    *(uint32_t *)(p+ 2) = __HYPERVISOR_iret;
+    *(uint16_t *)(p+ 6) = 0x82cd;  /* int  $0x82 */
+}
+
+
+/* --------------------------------------------------------------------- */
+/* called from assembler                                                 */
+
+asmlinkage void do_page_fault(struct regs_32 *regs)
+{
+    struct xen_cpu *cpu =get_cpu();
+    uint32_t cr2 = read_cr2();
+
+    vminfo.faults[XEN_FAULT_PAGE_FAULT]++;
+
+    if (context_is_emu(regs)) {
+        if (fixup_extable(regs)) {
+            return;
+        }
+        print_page_fault_info(0, cpu, regs, cr2);
+        panic("ring0 (emu) page fault", regs);
+    }
+
+    if (wrpt && regs->error == 3) {
+        /* kernel write to r/o page */
+        if (pf_fixup_readonly(regs, cr2)) {
+            return;
+        }
+    }
+
+    vminfo.faults[XEN_FAULT_PAGE_FAULT_GUEST]++;
+    bounce_trap(cpu, regs, 14, -1);
+}
-- 
1.6.0.2