Skip to contentSkip to conversation

Make BIOS irq0->inti2 override configurable from userspace

PATCH In review Beth Kon 2 participants 3 messages Apr 9, 2009 – Apr 16, 2009 KVM

lore
Share & tools
lore

Beth Kon

author
#0 lore raw

PATCH 1/2 Make BIOS irq0->inti2 override configurable from userspace

These patches resolve the irq0->inti2 override issue, and get the hpet working on kvm with and without -no-kvm-irqchip (i.e., when hpet takes over, it disables userspace or in-kernel pit as appropriate).

The irq0->inti2 override will always be used unless the kernel cannot do irq routing (i.e., compatibility with old kernels). So if the kernel is capable, userspace sets up irq0->inti2 via the irq routing interface, and adds the irq0->inti2 override to the MADT interrupt source override table, and the mp table (for the no-acpi case).

A couple of months ago, Marcelo was seeing RHEL5 guests complain of invalid checksum with these patches, but later he couldn't reproduce it, and I'm not seeing it now. While all guests still need to be fully tested, everything appears to be in order. I've tested on win2k864, win2k832, RHEL5.3 32 bit, and ubuntu 8.10 64 bit.

bios/rombios32.c +36 -4
index 4dea066..5cf1f54 100755
@@ -443,6 +443,7 @@ uint32_t cpuid_ext_features;
 unsigned long ram_size;
 uint64_t ram_end;
 uint8_t bios_uuid[16];
+uint8_t irq0_override;
 #ifdef BX_USE_EBDA_TABLES
 unsigned long ebda_cur_addr;
 #endif
@@ -475,6 +476,7 @@ void wrmsr_smp(uint32_t index, uint64_t val)
 #define QEMU_CFG_SIGNATURE  0x00
 #define QEMU_CFG_ID         0x01
 #define QEMU_CFG_UUID       0x02
+#define QEMU_CFG_IRQ0_OVERRIDE 0x0d
 
 int qemu_cfg_port;
 
@@ -516,6 +518,18 @@ void uuid_probe(void)
     memset(bios_uuid, 0, 16);
 }
 
+void irq0_override_probe(void)
+{
+#ifdef BX_QEMU
+    if(qemu_cfg_port) {
+        qemu_cfg_select(QEMU_CFG_IRQ0_OVERRIDE);
+        qemu_cfg_read(&irq0_override, 1);
+        return;
+    }
+#endif
+    memset(&irq0_override, 0, 1);
+}
+
 void cpu_probe(void)
 {
     uint32_t eax, ebx, ecx, edx;
@@ -1152,6 +1166,8 @@ static void mptable_init(void)
 
     /* irqs */
     for(i = 0; i < 16; i++) {
+        if (irq0_override && i == 2)
+            continue;
         putb(&q, 3); /* entry type = I/O interrupt */
         putb(&q, 0); /* interrupt type = vectored interrupt */
         putb(&q, 0); /* flags: po=0, el=0 */
@@ -1159,7 +1175,10 @@ static void mptable_init(void)
         putb(&q, 0); /* source bus ID = ISA */
         putb(&q, i); /* source bus IRQ */
         putb(&q, ioapic_id); /* dest I/O APIC ID */
-        putb(&q, i); /* dest I/O APIC interrupt in */
+        if (irq0_override && i == 0)
+            putb(&q, 2); /* dest I/O APIC interrupt in */
+        else
+            putb(&q, i); /* dest I/O APIC interrupt in */
     }
     /* patch length */
     len = q - mp_config_table;
@@ -1508,6 +1527,11 @@ void acpi_bios_init(void)
         sizeof(struct madt_processor_apic) * MAX_CPUS +
         sizeof(struct madt_io_apic);
     madt = (void *)(addr);
+    for (i = 0; i < 16; i++)
+        if (PCI_ISA_IRQ_MASK & (1U << i))
+            madt_size += sizeof(struct madt_intsrcovr);
+    if (irq0_override)
+        madt_size += sizeof(struct madt_intsrcovr);
     addr += madt_size;
 
     acpi_tables_size = addr - base_addr;
@@ -1597,8 +1621,15 @@ void acpi_bios_init(void)
         io_apic->interrupt = cpu_to_le32(0);
 
         intsrcovr = (struct madt_intsrcovr*)(io_apic + 1);
-        for ( i = 0; i < 16; i++ ) {
-            if ( PCI_ISA_IRQ_MASK & (1U << i) ) {
+        for (i = 0; i < 16; i++) {
+            if (irq0_override && i == 0) {
+                memset(intsrcovr, 0, sizeof(*intsrcovr));
+                intsrcovr->type   = APIC_XRUPT_OVERRIDE;
+                intsrcovr->length = sizeof(*intsrcovr);
+                intsrcovr->source = i;
+                intsrcovr->gsi    = 2;
+                intsrcovr->flags  = 0;  //conforms to bus specifications
+            } else if (PCI_ISA_IRQ_MASK & (1U << i)) {
                 memset(intsrcovr, 0, sizeof(*intsrcovr));
                 intsrcovr->type   = APIC_XRUPT_OVERRIDE;
                 intsrcovr->length = sizeof(*intsrcovr);
@@ -1610,7 +1641,6 @@ void acpi_bios_init(void)
                 continue;
             }
             intsrcovr++;
-            madt_size += sizeof(struct madt_intsrcovr);
         }
         acpi_build_table_header((struct acpi_table_header *)madt,
                                 "APIC", madt_size, 1);
@@ -2230,6 +2260,8 @@ void rombios32_init(uint32_t *s3_resume_vector, uint8_t *shutdown_flag)
 
     if (bios_table_cur_addr != 0) {
 
+        irq0_override_probe();
+
         mptable_init();
 
         uuid_probe();
qemu/hw/fw_cfg.c +1 -0
index e324e8d..f06dc3c 100644
@@ -279,6 +279,7 @@ void *fw_cfg_init(uint32_t ctl_port, uint32_t data_port,
     fw_cfg_add_bytes(s, FW_CFG_UUID, qemu_uuid, 16);
     fw_cfg_add_i16(s, FW_CFG_NOGRAPHIC, (uint16_t)nographic);
     fw_cfg_add_i16(s, FW_CFG_NB_CPUS, (uint16_t)smp_cpus);
+    fw_cfg_add_i16(s, FW_CFG_IRQ0_OVERRIDE, (uint16_t)irq0override);
 
     register_savevm("fw_cfg", -1, 1, fw_cfg_save, fw_cfg_load, s);
     qemu_register_reset(fw_cfg_reset, s);
qemu/hw/fw_cfg.h +1 -0
index 41a3dd0..ef6e8c6 100644
@@ -14,6 +14,7 @@
 #define FW_CFG_INITRD_ADDR      0x0a
 #define FW_CFG_INITRD_SIZE      0x0b
 #define FW_CFG_BOOT_DEVICE      0x0c
+#define FW_CFG_IRQ0_OVERRIDE    0x0d
 #define FW_CFG_MAX_ENTRY        0x10
 
 #define FW_CFG_WRITE_CHANNEL    0x4000
qemu/hw/ioapic.c +2 -3
index 0b70cf6..4dafb52 100644
@@ -23,6 +23,7 @@
 
 #include "hw.h"
 #include "pc.h"
+#include "sysemu.h"
 #include "qemu-timer.h"
 #include "host-utils.h"
 
@@ -95,14 +96,12 @@ void ioapic_set_irq(void *opaque, int vector, int level)
 {
     IOAPICState *s = opaque;
 
-#if 0
     /* ISA IRQs map to GSI 1-1 except for IRQ0 which maps
      * to GSI 2.  GSI maps to ioapic 1-1.  This is not
      * the cleanest way of doing it but it should work. */
 
-    if (vector == 0)
+    if (vector == 0 && irq0override)
         vector = 2;
-#endif
 
     if (vector >= 0 && vector < IOAPIC_NUM_PINS) {
         uint32_t mask = 1 << vector;
qemu/pc-bios/bios.bin binary +0 -0
index 04c1f86..c0b07b5 100644
Binary files a/qemu/pc-bios/bios.bin and b/qemu/pc-bios/bios.bin differ
qemu/qemu-kvm.c +4 -1
index 4164368..2755b5e 100644
@@ -812,7 +812,10 @@ int kvm_qemu_create_context(void)
                 return r;
         }
         for (i = 0; i < 24; ++i) {
-            r = kvm_add_irq_route(kvm_context, i, KVM_IRQCHIP_IOAPIC, i);
+            if (i == 0)
+                r = kvm_add_irq_route(kvm_context, i, KVM_IRQCHIP_IOAPIC, 2);
+            else if (i != 2)
+                r = kvm_add_irq_route(kvm_context, i, KVM_IRQCHIP_IOAPIC, i);
             if (r < 0)
                 return r;
         }
qemu/sysemu.h +1 -0
index d765465..fd127bd 100644
@@ -94,6 +94,7 @@ extern int graphic_width;
 extern int graphic_height;
 extern int graphic_depth;
 extern int nographic;
+extern int irq0override;
 extern const char *keyboard_layout;
 extern int win2k_install_hack;
 extern int rtc_td_hack;
qemu/vl.c +8 -2
index b3da7ad..b982b53 100644
@@ -205,6 +205,7 @@ static int vga_ram_size;
 enum vga_retrace_method vga_retrace_method = VGA_RETRACE_DUMB;
 static DisplayState *display_state;
 int nographic;
+int irq0override;
 static int curses;
 static int sdl;
 const char* keyboard_layout = NULL;
@@ -4516,6 +4517,7 @@ int main(int argc, char **argv, char **envp)
 #endif
     snapshot = 0;
     nographic = 0;
+    irq0override = 1;
     curses = 0;
     kernel_filename = NULL;
     kernel_cmdline = "";
@@ -5536,8 +5538,12 @@ int main(int argc, char **argv, char **envp)
         }
     }
 
-    if (kvm_enabled())
-	kvm_init_ap();
+    if (kvm_enabled()) {
+       kvm_init_ap();
+        if (kvm_irqchip && !kvm_has_gsi_routing(kvm_context)) {
+           irq0override = 0;
+        }
+    }
 
     machine->init(ram_size, vga_ram_size, boot_devices,
                   kernel_filename, kernel_cmdline, initrd_filename, cpu_model);
author
#1 lore raw

PATCH 2/2 Finish HPET implementation for KVM

bios/acpi-dsdt.dsl +18 -0
index 06ab25d..84697db 100755
@@ -307,6 +307,24 @@ DefinitionBlock (
                     ,, , AddressRangeMemory, TypeStatic)
             })
         }
+        Device(HPET) {
+            Name(_HID,  EISAID("PNP0103"))
+            Name(_UID, 0)
+            Method (_STA, 0, NotSerialized) {
+                    Return(0x0F)
+            }
+            Name(_CRS, ResourceTemplate() {
+                DWordMemory(
+                    ResourceConsumer, PosDecode, MinFixed, MaxFixed,
+                    NonCacheable, ReadWrite,
+                    0x00000000,
+                    0xFED00000,
+                    0xFED003FF,
+                    0x00000000,
+                    0x00000400 /* 1K memory: FED00000 - FED003FF */
+                )
+            })
+        }
     }
 
     Scope(\_SB.PCI0) {
bios/rombios32.c +43 -1

80 lines not on this page · raw message

qemu/hw/hpet.c +41 -3

69 lines not on this page · raw message

qemu/vl.c +6 -1
index b982b53..d662fba 100644
@@ -5539,9 +5539,14 @@ int main(int argc, char **argv, char **envp)
     }
 
     if (kvm_enabled()) {
-       kvm_init_ap();
+        kvm_init_ap();
         if (kvm_irqchip && !kvm_has_gsi_routing(kvm_context)) {
            irq0override = 0;
+           /* if kernel can't do irq routing, interrupt source
+            * override 0->2 can not be set up as required by hpet,
+            * so disable hpet.
+            */
+           no_hpet=1;
         }
     }
 
#2 lore raw
In reply to Beth Kon
Quoted 8 lines · Beth Kon

These patches resolve the irq0->inti2 override issue, and get the hpet working
on kvm with and without -no-kvm-irqchip (i.e., when hpet takes over, it disables userspace or in-kernel pit as appropriate).

The irq0->inti2 override will always be used unless the kernel cannot do irq routing (i.e., compatibility with old kernels). So if the kernel is capable, userspace sets up irq0->inti2 via the irq routing interface, and adds the irq0->inti2 override to the MADT interrupt source override table, and the mp table (for the no-acpi case).

A couple of months ago, Marcelo was seeing RHEL5 guests complain of invalid
checksum with these patches, but later he couldn't reproduce it, and I'm not seeing it now. While all guests still need to be fully tested, everything appears to be in order. I've tested on win2k864, win2k832, RHEL5.3 32 bit, and ubuntu 8.10 64 bit.

Both patches look good. However, the bios in kvm-userspace.git is pretty old. To ease later merging, I'd like to update the bios first, so please rebase the patches once I've done that. It will take a few days since large merges are blocked while I struggle to get a working kvm-85 out.

Please also separate the bios patches from qemu patches.

Signature · 1 line
Do not meddle in the internals of kernels, for they are subtle and quick to panic.