diff --git a/share/man/man4/pv.4 b/share/man/man4/pv.4
new file mode 100644
index 00000000000..f25ba4a2475
--- /dev/null
+++ b/share/man/man4/pv.4
@@ -0,0 +1,44 @@
+.\" $NetBSD$
+.\"
+.\" Copyright (c) 2024 The NetBSD Foundation, Inc.
+.\" All rights reserved.
+.\"
+.\" Redistribution and use in source and binary forms, with or without
+.\" modification, are permitted provided that the following conditions
+.\" are met:
+.\" 1. Redistributions of source code must retain the above copyright
+.\" notice, this list of conditions and the following disclaimer.
+.\" 2. Redistributions in binary form must reproduce the above copyright
+.\" notice, this list of conditions and the following disclaimer in the
+.\" documentation and/or other materials provided with the distribution.
+.\"
+.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+.\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+.\" POSSIBILITY OF SUCH DAMAGE.
+.\"
+.\" ported from OpenBSD
+.Dd January 2024
+.Dt PV 4
+.Os
+.Sh NAME
+.Nm pv
+.Nd paravirtual device tree root
+.Sh SYNOPSIS
+.Cd "pv* at pvbus?"
+.Sh DESCRIPTION
+.Nm
+driver is used on virtual machines that are running on hypervisors.
+It provides a pseudo-bus for all paravirtual devices that do not
+attach to a well-known bus like
+.Xr pci 4 .
+.Sh SEE ALSO
+.Xr virtio 4 ,
+.Xr mmio_virtio 4
diff --git a/share/man/man4/virtio_mmio.4 b/share/man/man4/virtio_mmio.4
new file mode 100644
index 00000000000..6f92cb3fdc4
--- /dev/null
+++ b/share/man/man4/virtio_mmio.4
@@ -0,0 +1,75 @@
+.\" $NetBSD$
+.\"
+.\" Copyright (c) 2024 The NetBSD Foundation, Inc.
+.\" All rights reserved.
+.\"
+.\" Redistribution and use in source and binary forms, with or without
+.\" modification, are permitted provided that the following conditions
+.\" are met:
+.\" 1. Redistributions of source code must retain the above copyright
+.\" notice, this list of conditions and the following disclaimer.
+.\" 2. Redistributions in binary form must reproduce the above copyright
+.\" notice, this list of conditions and the following disclaimer in the
+.\" documentation and/or other materials provided with the distribution.
+.\"
+.\" THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+.\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+.\" TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+.\" PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+.\" BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+.\" CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+.\" SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+.\" INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+.\" CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+.\" ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+.\" POSSIBILITY OF SUCH DAMAGE.
+.\"
+.Dd January 2024
+.Dt VIRTIO_MMIO 4
+.Os NetBSD
+.Sh NAME
+.Nm virtio_mmio
+.Nd VirtIO over memory mapped device.
+.Sh SYNOPSIS
+.Cd "pv* at pvbus?"
+.Cd "virtio* at pv?"
+.Pp
+.Cd "acpi0 at mainbus0"
+.Cd "virtio* at acpi?"
+.Sh DESCRIPTION
+.Nm
+can be used in virtual environments without
+.Xr pci 4
+support (a common situation in embedded devices models) might use simple
+memory mapped device (
+.Nm
+) instead of the
+.Xr pci 4
+device.
+.Pp
+The memory mapped
+.Xr virtio 4
+device behaviour is based on the
+.Xr pci 4
+device specification. Therefore most operations including device initialization,
+queues configuration and buffer transfers are nearly identical.
+.Pp
+Unlike
+.Xr pci 4
+,
+.Nm
+provides no generic device discovery mechanism. For each device, the guest OS will
+need to know the location of the registers and interrupt(s) used.
+.Pp
+Device location can be read from either
+.Xr acpi 4
+or via kernel command line parameters, implemented as a
+.Xr pv 4
+virtual device.
+.Sh SEE ALSO
+.Xr virtio 4
+.Pp
+.Rs
+.%T Virtual I/O Device (VIRTIO) Version 1.2
+.%U https://docs.oasis-open.org/virtio/virtio/v1.2/virtio-v1.2.html
+.Re
diff --git a/sys/arch/amd64/amd64/amd64_mainbus.c b/sys/arch/amd64/amd64/amd64_mainbus.c
index 8a1bb0c55c6..f9d1118a54f 100644
--- a/sys/arch/amd64/amd64/amd64_mainbus.c
+++ b/sys/arch/amd64/amd64/amd64_mainbus.c
@@ -50,6 +50,7 @@ __KERNEL_RCSID(0, "$NetBSD: amd64_mainbus.c,v 1.7 2021/08/07 16:18:41 thorpej Ex
#include "isadma.h"
#include "acpica.h"
#include "ipmi.h"
+#include "pvbus.h"
#include "opt_acpi.h"
#include "opt_mpbios.h"
@@ -79,6 +80,9 @@ __KERNEL_RCSID(0, "$NetBSD: amd64_mainbus.c,v 1.7 2021/08/07 16:18:41 thorpej Ex
#include <arch/x86/pci/msipic.h>
#endif /* __HAVE_PCI_MSI_MSIX */
#endif
+#if NPVBUS > 0
+#include <dev/pv/pvvar.h>
+#endif
/*
* XXXfvdl ACPI
@@ -100,6 +104,9 @@ union amd64_mainbus_attach_args {
#if NIPMI > 0
struct ipmi_attach_args mba_ipmi;
#endif
+#if NPVBUS > 0
+ struct pvbus_attach_args mba_pvba;
+#endif
};
/*
@@ -155,7 +162,7 @@ amd64_mainbus_match(device_t parent, cfdata_t match, void *aux)
void
amd64_mainbus_attach(device_t parent, device_t self, void *aux)
{
-#if NISA > 0 || NPCI > 0 || NACPICA > 0 || NIPMI > 0
+#if NISA > 0 || NPCI > 0 || NACPICA > 0 || NIPMI > 0 || NPVBUS > 0
union amd64_mainbus_attach_args mba;
#endif
@@ -237,6 +244,12 @@ amd64_mainbus_attach(device_t parent, device_t self, void *aux)
}
#endif
+#if NPVBUS > 0
+ mba.mba_pvba.pvba_busname = "pvbus";
+ config_found(self, &mba.mba_pvba.pvba_busname, NULL,
+ CFARGS(.iattr = "pvbus"));
+#endif
+
if (!pmf_device_register(self, NULL, NULL))
aprint_error_dev(self, "couldn't establish power handler\n");
}
diff --git a/sys/arch/amd64/amd64/genassym.cf b/sys/arch/amd64/amd64/genassym.cf
index b5c7330a229..facd9216f75 100644
--- a/sys/arch/amd64/amd64/genassym.cf
+++ b/sys/arch/amd64/amd64/genassym.cf
@@ -373,6 +373,7 @@ define BST_TYPE offsetof(struct bus_space_tag, bst_type)
define VM_GUEST_XENPV VM_GUEST_XENPV
define VM_GUEST_XENPVH VM_GUEST_XENPVH
+define VM_GUEST_GENPVH VM_GUEST_GENPVH
ifdef XEN
define CPU_INFO_VCPU offsetof(struct cpu_info, ci_vcpu)
@@ -381,6 +382,12 @@ define SIR_XENIPL_VM SIR_XENIPL_VM
define SIR_XENIPL_SCHED SIR_XENIPL_SCHED
define SIR_XENIPL_HIGH SIR_XENIPL_HIGH
define EVTCHN_UPCALL_MASK offsetof(struct vcpu_info, evtchn_upcall_mask)
+define HVM_START_INFO_SIZE sizeof(struct hvm_start_info)
+define START_INFO_VERSION offsetof(struct hvm_start_info, version)
+define MMAP_PADDR offsetof(struct hvm_start_info, memmap_paddr)
+define MMAP_ENTRIES offsetof(struct hvm_start_info, memmap_entries)
+define MMAP_ENTRY_SIZE sizeof(struct hvm_memmap_table_entry)
+define CMDLINE_PADDR offsetof(struct hvm_start_info, cmdline_paddr)
ifdef XENPV
define XEN_PT_BASE offsetof(struct start_info, pt_base)
define XEN_NR_PT_FRAMES offsetof(struct start_info, nr_pt_frames)
diff --git a/sys/arch/amd64/amd64/locore.S b/sys/arch/amd64/amd64/locore.S
index 089c7388e18..97519bcb6a1 100644
--- a/sys/arch/amd64/amd64/locore.S
+++ b/sys/arch/amd64/amd64/locore.S
@@ -278,11 +278,12 @@
#ifdef XEN
#define __ASSEMBLY__
+#include <xen/include/public/arch-x86/cpuid.h>
#include <xen/include/public/elfnote.h>
#include <xen/include/public/xen.h>
#define ELFNOTE(name, type, desctype, descdata...) \
-.pushsection .note.name ; \
+.pushsection .note.name, "a", @note ; \
.align 4 ; \
.long 2f - 1f /* namesz */ ; \
.long 4f - 3f /* descsz */ ; \
@@ -306,7 +307,7 @@
ELFNOTE(Xen, XEN_ELFNOTE_ENTRY, .quad, start)
#else
ELFNOTE(Xen, XEN_ELFNOTE_PADDR_OFFSET, .quad, 0)
- ELFNOTE(Xen, XEN_ELFNOTE_PHYS32_ENTRY, .long, RELOC(start_xen32))
+ ELFNOTE(Xen, XEN_ELFNOTE_PHYS32_ENTRY, .long, RELOC(start_genpvh))
#endif /* XENPV */
ELFNOTE(Xen, XEN_ELFNOTE_HYPERCALL_PAGE, .quad, hypercall_page)
ELFNOTE(Xen, XEN_ELFNOTE_HV_START_LOW, .quad, HYPERVISOR_VIRT_START)
@@ -973,7 +974,6 @@ longmode_hi:
movl %eax,_C_LABEL(cpuid_level)
movl $VM_GUEST_XENPV, _C_LABEL(vm_guest)
-
/*
* Initialize cpu_info_primary.ci_self := &cpu_info_primary,
* and initialize some MSRs with
@@ -1035,7 +1035,7 @@ END(start)
# if !defined(XENPV)
/* entry point for Xen PVH */
.code32
-ENTRY(start_xen32)
+ENTRY(start_genpvh)
/* Xen doesn't start us with a valid gdt */
movl $RELOC(gdtdesc32), %eax
lgdt (%eax)
@@ -1051,13 +1051,98 @@ ENTRY(start_xen32)
movl $RELOC(tmpstk),%esp
/* clear BSS */
- xorl %eax,%eax
+ xorl %eax,%eax
movl $RELOC(__bss_start),%edi
movl $RELOC(_end),%ecx
subl %edi,%ecx
rep
stosb
+ /*
+ * Here, we have 2 cases :
+ *
+ * 1) We have been started by Xen
+ * 2) We have been started by another VMM (Qemu, Firecracker, ...)
+ *
+ * The main difference is that, when we are started by Xen,
+ * %ebx (addr of the hvm_start_info structure) is pointing to a
+ * location that will be mapped correctly later.
+ *
+ * In the second case, we have to copy this structure (and all
+ * the information contained in it) to a location that will be
+ * mapped later : __kernel_end
+ *
+ * To distinguish between the 2 cases, we'll use the 'cpuid' instruction
+ */
+
+ push %ebx
+ xorl %eax, %eax
+ cpuid
+ cmpl $0x1, %eax /* Check if we can call CPUID with eax=1 */
+ jb .start_genpvh
+ xorl %eax, %eax
+ inc %eax
+ cpuid
+ shr $31, %ecx
+ testb $1, %cl /* Check if bit 31 of ECX (hypervisor) is set */
+ jz .start_genpvh
+ xorl %eax, %eax
+ inc %eax
+ shl $30, %eax
+ cpuid /* Calling cpuid with eax=0x40000000 */
+ cmp $XEN_CPUID_SIGNATURE_EBX, %ebx /* "VneX" */
+ je .start_xen32
+
+ /* We have been started by a VMM that is *not* Xen */
+
+.start_genpvh:
+
+ /* First, copy the hvm_start_info structure to __kernel_end */
+ pop %ebx
+ movl %ebx, %esi
+ movl $RELOC(__kernel_end), %edi
+ movl $HVM_START_INFO_SIZE, %ecx
+ shrl $2, %ecx
+ rep movsl
+
+ /* Copy cmdline_paddr after hvm_start_info */
+ movl CMDLINE_PADDR(%ebx), %esi
+ movl $RELOC(__kernel_end), %ecx
+ movl %edi, CMDLINE_PADDR(%ecx) /* Set new cmdline_paddr in hvm_start_info */
+ .cmdline_copy:
+ movb (%esi), %al
+ movsb
+ cmp $0, %al
+ jne .cmdline_copy
+
+ /* Copy memmap_paddr after cmdline (only if hvm_start_info->version != 0) */
+ xorl %eax, %eax
+ cmpl START_INFO_VERSION(%ebx), %eax
+ je .reload_ebx
+ movl MMAP_PADDR(%ebx), %esi
+ movl $RELOC(__kernel_end), %ecx
+ movl %edi, MMAP_PADDR(%ecx) /* Set new memmap_paddr in hvm_start_info */
+ movl MMAP_ENTRIES(%ebx), %eax /* Get memmap_entries */
+ movl $MMAP_ENTRY_SIZE, %ebx
+ mull %ebx /* eax * ebx => edx:eax */
+ movl %eax, %ecx
+ shll $2, %ecx
+ rep movsl
+
+.reload_ebx:
+ movl $RELOC(__kernel_end), %ebx
+
+ /* announce ourself */
+ movl $VM_GUEST_GENPVH, RELOC(vm_guest)
+
+ jmp .save_hvm_start_paddr
+
+.start_xen32:
+ pop %ebx
+ movl $VM_GUEST_XENPVH, RELOC(vm_guest)
+
+.save_hvm_start_paddr:
+
/*
* save addr of the hvm_start_info structure. This is also the end
* of the symbol table
@@ -1069,6 +1154,9 @@ ENTRY(start_xen32)
movl %eax,(%ebp)
movl $KERNBASE_HI,4(%ebp)
/* get a page for HYPERVISOR_shared_info */
+ /* this is only needed if we are running on Xen */
+ cmpl $VM_GUEST_XENPVH, RELOC(vm_guest)
+ jne .add_hvm_start_info_page
addl $PAGE_SIZE, %ebx
addl $PGOFSET,%ebx
andl $~PGOFSET,%ebx
@@ -1076,6 +1164,7 @@ ENTRY(start_xen32)
movl %ebx,(%ebp)
movl $0,4(%ebp)
/* XXX assume hvm_start_info+dependant structure fits in a single page */
+.add_hvm_start_info_page:
addl $PAGE_SIZE, %ebx
addl $PGOFSET,%ebx
andl $~PGOFSET,%ebx
@@ -1084,10 +1173,8 @@ ENTRY(start_xen32)
movl %ebx,(%ebp)
movl $KERNBASE_HI,4(%ebp)
- /* announce ourself */
- movl $VM_GUEST_XENPVH, RELOC(vm_guest)
jmp .Lbiosbasemem_finished
-END(start_xen32)
+END(start_genpvh)
.code64
# endif /* !XENPV */
/* space for the hypercall call page */
@@ -1099,6 +1186,7 @@ ENTRY(hypercall_page) /* Returns -1, on HYPERVISOR_xen_version() */
retq
.align HYPERCALL_PAGE_OFFSET, 0x90
END(hypercall_page)
+
#endif /* XEN */
/*
diff --git a/sys/arch/amd64/amd64/machdep.c b/sys/arch/amd64/amd64/machdep.c
index 375d7a41e7f..f1c8f5c66a6 100644
--- a/sys/arch/amd64/amd64/machdep.c
+++ b/sys/arch/amd64/amd64/machdep.c
@@ -1525,8 +1525,10 @@ init_x86_64_ksyms(void)
} else {
uintptr_t endp = (uintptr_t)(void *)&end;
- ksyms_addsyms_elf(*(long *)endp,
- ((long *)endp) + 1, esym);
+ if (vm_guest == VM_GUEST_GENPVH)
+ ksyms_addsyms_elf(0, ((long *)endp) + 1, esym);
+ else
+ ksyms_addsyms_elf(*(long *)endp, ((long *)endp) + 1, esym);
}
#endif
}
@@ -1710,7 +1712,7 @@ init_x86_64(paddr_t first_avail)
#endif
#ifdef XEN
- if (vm_guest == VM_GUEST_XENPVH)
+ if (vm_guest == VM_GUEST_XENPVH || vm_guest == VM_GUEST_GENPVH)
xen_parse_cmdline(XEN_PARSE_BOOTFLAGS, NULL);
#endif
init_pte();
diff --git a/sys/arch/amd64/conf/files.amd64 b/sys/arch/amd64/conf/files.amd64
index 99e0e512938..d4e2fc5f38b 100644
--- a/sys/arch/amd64/conf/files.amd64
+++ b/sys/arch/amd64/conf/files.amd64
@@ -94,7 +94,7 @@ include "dev/i2o/files.i2o"
# XXX BIOS32 only if something that uses it is configured!
device mainbus: isabus, pcibus, bios32, acpibus, cpubus, ioapicbus,
- ipmibus, hypervisorbus
+ ipmibus, hypervisorbus, pvbus
attach mainbus at root
file arch/amd64/amd64/amd64_mainbus.c mainbus & !xenpv
file arch/x86/x86/mainbus.c mainbus
@@ -200,4 +200,7 @@ file dev/acpi/vmbus_acpi.c vmbus_acpi
# VMEbus support
include "dev/vme/files.vme"
+# PVbus support
+include "dev/pv/files.pv"
+
include "arch/amd64/conf/majors.amd64"
diff --git a/sys/arch/x86/acpi/acpi_machdep.c b/sys/arch/x86/acpi/acpi_machdep.c
index 9dbfe49ca5f..b75767fe615 100644
--- a/sys/arch/x86/acpi/acpi_machdep.c
+++ b/sys/arch/x86/acpi/acpi_machdep.c
@@ -158,7 +158,7 @@ out:
}
#else
#ifdef XEN
- if (vm_guest == VM_GUEST_XENPVH) {
+ if (vm_guest == VM_GUEST_XENPVH || vm_guest == VM_GUEST_GENPVH) {
PhysicalAddress = hvm_start_info->rsdp_paddr;
if (PhysicalAddress)
return PhysicalAddress;
diff --git a/sys/arch/x86/include/cpu.h b/sys/arch/x86/include/cpu.h
index 0dae029202a..c82da831157 100644
--- a/sys/arch/x86/include/cpu.h
+++ b/sys/arch/x86/include/cpu.h
@@ -516,6 +516,7 @@ typedef enum vm_guest {
VM_GUEST_VMWARE,
VM_GUEST_KVM,
VM_GUEST_VIRTUALBOX,
+ VM_GUEST_GENPVH,
VM_LAST
} vm_guest_t;
extern vm_guest_t vm_guest;
diff --git a/sys/arch/x86/x86/bus_dma.c b/sys/arch/x86/x86/bus_dma.c
index b47b4b1bc2e..7d2753bb7c2 100644
--- a/sys/arch/x86/x86/bus_dma.c
+++ b/sys/arch/x86/x86/bus_dma.c
@@ -319,7 +319,7 @@ _bus_dmamap_create(bus_dma_tag_t t, bus_size_t size, int nsegments,
goto out;
}
- if (map->_dm_bounce_thresh != 0)
+ if (map->_dm_bounce_thresh != 0 || map->_dm_segcnt == 1)
cookieflags |= X86_DMA_MIGHT_NEED_BOUNCE;
if ((cookieflags & X86_DMA_MIGHT_NEED_BOUNCE) == 0) {
diff --git a/sys/arch/x86/x86/consinit.c b/sys/arch/x86/x86/consinit.c
index 7a65f1bfa09..f72f872ab54 100644
--- a/sys/arch/x86/x86/consinit.c
+++ b/sys/arch/x86/x86/consinit.c
@@ -171,6 +171,7 @@ consinit(void)
#if (NCOM > 0)
int rv;
#endif
+ char console_devname[16] = "";
#ifdef XENPVHVM
if (vm_guest == VM_GUEST_XENPVH) {
@@ -178,6 +179,13 @@ consinit(void)
return;
/* fallback to native console selection, usefull for dom0 PVH */
}
+ if (vm_guest == VM_GUEST_GENPVH) {
+ union xen_cmdline_parseinfo xcp;
+ /* get console= parameter from generic PVH VMM */
+ xen_parse_cmdline(XEN_PARSE_CONSOLE, &xcp);
+ strncpy(console_devname, xcp.xcp_console,
+ sizeof(console_devname));
+ }
#endif
if (initted)
return;
@@ -188,7 +196,10 @@ consinit(void)
if (!consinfo)
#endif
consinfo = &default_consinfo;
-
+ /* console= parameter was not passed via a generic PVH VMM */
+ if (!console_devname[0])
+ strncpy(console_devname, consinfo->devname,
+ sizeof(console_devname));
#if (NGENFB > 0)
#if defined(XENPVHVM) && defined(DOM0OPS)
if (vm_guest == VM_GUEST_XENPVH && xendomain_is_dom0())
@@ -197,8 +208,7 @@ consinit(void)
#endif /* XENPVHVM */
fbinfo = lookup_bootinfo(BTINFO_FRAMEBUFFER);
#endif
-
- if (!strcmp(consinfo->devname, "pc")) {
+ if (!strcmp(console_devname, "pc")) {
int error;
#if (NGENFB > 0)
if (fbinfo && fbinfo->physaddr > 0) {
@@ -254,7 +264,7 @@ dokbd:
return;
}
#if (NCOM > 0)
- if (!strcmp(consinfo->devname, "com")) {
+ if (!strcmp(console_devname, "com")) {
int addr = consinfo->addr;
int speed = consinfo->speed;
@@ -278,14 +288,14 @@ dokbd:
}
#endif
#if (NNULLCONS > 0)
- if (!strcmp(consinfo->devname, "nullcons")) {
+ if (!strcmp(console_devname, "nullcons")) {
void nullcninit(struct consdev *cn);
nullcninit(0);
return;
}
#endif
- panic("invalid console device %s", consinfo->devname);
+ panic("invalid console device %s", console_devname);
}
#ifdef KGDB
diff --git a/sys/arch/x86/x86/identcpu.c b/sys/arch/x86/x86/identcpu.c
index ec2fa0620b7..9856cf9fb15 100644
--- a/sys/arch/x86/x86/identcpu.c
+++ b/sys/arch/x86/x86/identcpu.c
@@ -1044,6 +1044,7 @@ static const struct vm_name_guest vm_bios_vendors[] = {
{ "BHYVE", VM_GUEST_VM }, /* bhyve */
{ "Seabios", VM_GUEST_VM }, /* KVM */
{ "innotek GmbH", VM_GUEST_VIRTUALBOX }, /* Oracle VirtualBox */
+ { "Generic PVH", VM_GUEST_GENPVH}, /* Generic PVH */
};
static const struct vm_name_guest vm_system_products[] = {
@@ -1065,6 +1066,7 @@ identify_hypervisor(void)
switch (vm_guest) {
case VM_GUEST_XENPV:
case VM_GUEST_XENPVH:
+ case VM_GUEST_GENPVH:
/* guest type already known, no bios info */
return;
default:
diff --git a/sys/arch/x86/x86/mpbios.c b/sys/arch/x86/x86/mpbios.c
index a3565a44251..85b3e38e3e2 100644
--- a/sys/arch/x86/x86/mpbios.c
+++ b/sys/arch/x86/x86/mpbios.c
@@ -209,6 +209,9 @@ static void mpbios_int(const uint8_t *, int, struct mp_intr_map *);
static const void *mpbios_map(paddr_t, int, struct mp_map *);
static void mpbios_unmap(struct mp_map *);
+#ifdef MPTABLE_LINUX_BUG_COMPAT
+static uint16_t compute_entry_count(const uint8_t *, const uint8_t *);
+#endif
/*
* globals to help us bounce our way through parsing the config table.
*/
@@ -333,6 +336,17 @@ mpbios_probe(device_t self)
if (mp_fps != NULL)
goto found;
+ /*
+ * Linux assumes that it always has 640 kB of base memory and
+ * searches for the MP table at 639k regardless of whether that
+ * address is present in the system memory map. Some VM systems
+ * rely on this buggy behaviour.
+ */
+ mp_fps = mpbios_search(self, 639 * 1024, 1024 / 4, &mp_fp_map);
+ if (mp_fps != NULL)
+ goto found;
+
+
/* nothing found */
return 0;
@@ -533,6 +547,31 @@ static const uint8_t dflt_lint_tab[2] = {
};
+#ifdef MPTABLE_LINUX_BUG_COMPAT
+/* Compute the correct entry_count value. */
+static uint16_t
+compute_entry_count(const uint8_t *entry, const uint8_t *end)
+{
+ size_t nentries = 0;
+
+ while (entry < end) {
+ switch (*entry) {
+ case MPS_MCT_CPU:
+ case MPS_MCT_BUS:
+ case MPS_MCT_IOAPIC:
+ case MPS_MCT_IOINT:
+ case MPS_MCT_LINT:
+ break;
+ default:
+ panic("%s: Unknown MP Config Entry %d\n", __func__,
+ (int)*entry);
+ }
+ entry += mp_conf[*entry].length;;
+ nentries++;
+ }
+ return (uint16_t)(nentries);
+}
+#endif
/*
* 1st pass on BIOS's Intel MP specification table.
*
@@ -557,6 +596,9 @@ mpbios_scan(device_t self, int *ncpup)
int intr_cnt, cur_intr;
#if NLAPIC > 0
paddr_t lapic_base;
+#endif
+#ifdef MPTABLE_LINUX_BUG_COMPAT
+ uint16_t countfix = 0;
#endif
const struct dflt_conf_entry *dflt_conf;
const int *dflt_bus_irq;
@@ -677,6 +719,13 @@ mpbios_scan(device_t self, int *ncpup)
position += sizeof(*mp_cth);
count = mp_cth->entry_count;
+#ifdef MPTABLE_LINUX_BUG_COMPAT
+ if (count == 0) {
+ /* count the correct entry_count */
+ countfix = compute_entry_count(position, end);
+ count = countfix;
+ }
+#endif
intr_cnt = 0;
while ((count--) && (position < end)) {
@@ -721,6 +770,10 @@ mpbios_scan(device_t self, int *ncpup)
/* re-walk the table, recording info of interest */
position = (const uint8_t *)mp_cth + sizeof(*mp_cth);
count = mp_cth->entry_count;
+#ifdef MPTABLE_LINUX_BUG_COMPAT
+ if (count == 0)
+ count = countfix;
+#endif
cur_intr = 0;
while ((count--) && (position < end)) {
diff --git a/sys/arch/x86/x86/x86_autoconf.c b/sys/arch/x86/x86/x86_autoconf.c
index fd8688fe9cc..ba4b698d067 100644
--- a/sys/arch/x86/x86/x86_autoconf.c
+++ b/sys/arch/x86/x86/x86_autoconf.c
@@ -540,7 +540,7 @@ void
cpu_bootconf(void)
{
#ifdef XEN
- if (vm_guest == VM_GUEST_XENPVH) {
+ if (vm_guest == VM_GUEST_XENPVH || vm_guest == VM_GUEST_GENPVH) {
xen_bootconf();
return;
}
diff --git a/sys/arch/x86/x86/x86_machdep.c b/sys/arch/x86/x86/x86_machdep.c
index fabd76b1155..8a2364b6ecc 100644
--- a/sys/arch/x86/x86/x86_machdep.c
+++ b/sys/arch/x86/x86/x86_machdep.c
@@ -911,7 +911,7 @@ init_x86_clusters(void)
* the boot program).
*/
#ifdef XEN
- if (vm_guest == VM_GUEST_XENPVH) {
+ if (vm_guest == VM_GUEST_XENPVH || vm_guest == VM_GUEST_GENPVH) {
x86_add_xen_clusters();
}
#endif /* XEN */
diff --git a/sys/arch/xen/xen/hypervisor.c b/sys/arch/xen/xen/hypervisor.c
index 47b7dc3ec0a..7e26b7696a6 100644
--- a/sys/arch/xen/xen/hypervisor.c
+++ b/sys/arch/xen/xen/hypervisor.c
@@ -241,10 +241,25 @@ void
init_xen_early(void)
{
const char *cmd_line;
+ if (vm_guest != VM_GUEST_XENPVH && vm_guest != VM_GUEST_GENPVH)
+ return;
+
+ hvm_start_info = (void *)((uintptr_t)hvm_start_paddr + KERNBASE);
+
+ if (hvm_start_info->cmdline_paddr != 0) {
+ cmd_line =
+ (void *)((uintptr_t)hvm_start_info->cmdline_paddr + KERNBASE);
+ strlcpy(xen_start_info.cmd_line, cmd_line,
+ sizeof(xen_start_info.cmd_line));
+ } else {
+ xen_start_info.cmd_line[0] = '\0';
+ }
+ xen_start_info.flags = hvm_start_info->flags;
+
if (vm_guest != VM_GUEST_XENPVH)
return;
+
xen_init_hypercall_page();
- hvm_start_info = (void *)((uintptr_t)hvm_start_paddr + KERNBASE);
HYPERVISOR_shared_info = (void *)((uintptr_t)HYPERVISOR_shared_info_pa + KERNBASE);
struct xen_add_to_physmap xmap = {
@@ -262,15 +277,6 @@ init_xen_early(void)
}
delay_func = x86_delay = xen_delay;
x86_initclock_func = xen_initclocks;
- if (hvm_start_info->cmdline_paddr != 0) {
- cmd_line =
- (void *)((uintptr_t)hvm_start_info->cmdline_paddr + KERNBASE);
- strlcpy(xen_start_info.cmd_line, cmd_line,
- sizeof(xen_start_info.cmd_line));
- } else {
- xen_start_info.cmd_line[0] = '\0';
- }
- xen_start_info.flags = hvm_start_info->flags;
}
diff --git a/sys/dev/acpi/files.acpi b/sys/dev/acpi/files.acpi
index 6763f983669..9f87bb3b638 100644
--- a/sys/dev/acpi/files.acpi
+++ b/sys/dev/acpi/files.acpi
@@ -273,7 +273,7 @@ attach amdccp at acpinodebus with amdccp_acpi
file dev/acpi/amdccp_acpi.c amdccp_acpi
# QEMU Virtio
-attach virtio at acpinodebus with virtio_acpi
+attach virtio at acpinodebus with virtio_acpi: virtio_mmio
file dev/acpi/virtio_acpi.c virtio_acpi
# OHCI-compliant USB controller
diff --git a/sys/dev/ldvar.h b/sys/dev/ldvar.h
index 0aa0c097a74..5b9d8bc2075 100644
--- a/sys/dev/ldvar.h
+++ b/sys/dev/ldvar.h
@@ -56,6 +56,7 @@ struct ld_softc {
uint64_t sc_secperunit; /* # sectors in total */
int sc_secsize; /* sector size in bytes */
int sc_maxxfer; /* max xfer size in bytes */
+ int sc_maxnsegs; /* maximum number of segments */
int sc_maxqueuecnt; /* maximum h/w queue depth */
char *sc_typename; /* inquiry data */
diff --git a/sys/dev/pci/ld_virtio.c b/sys/dev/pci/ld_virtio.c
index 0cea794c9a0..afdc1e48aa8 100644
--- a/sys/dev/pci/ld_virtio.c
+++ b/sys/dev/pci/ld_virtio.c
@@ -196,6 +196,7 @@ ld_virtio_alloc_reqs(struct ld_virtio_softc *sc, int qsize)
memset(vaddr, 0, allocsize);
for (i = 0; i < qsize; i++) {
struct virtio_blk_req *vr = &sc->sc_reqs[i];
+ int nsegs;
r = bus_dmamap_create(virtio_dmat(sc->sc_virtio),
offsetof(struct virtio_blk_req, vr_bp),
1,
@@ -219,10 +220,18 @@ ld_virtio_alloc_reqs(struct ld_virtio_softc *sc, int qsize)
"error code %d\n", r);
goto err_reqs;
}
+ /*
+ * if d->sc_maxnsegs == VIRTIO_BLK_MIN_SEGMENTS + 1, the
+ * device only supports a single data segment.
+ */
+ if (ld->sc_maxnsegs == VIRTIO_BLK_MIN_SEGMENTS + 1)
+ nsegs = ld->sc_maxnsegs - VIRTIO_BLK_MIN_SEGMENTS;
+ else
+ nsegs = (ld->sc_maxxfer / NBPG) + VIRTIO_BLK_MIN_SEGMENTS;
+
r = bus_dmamap_create(virtio_dmat(sc->sc_virtio),
ld->sc_maxxfer,
- (ld->sc_maxxfer / NBPG) +
- VIRTIO_BLK_MIN_SEGMENTS,
+ nsegs,
ld->sc_maxxfer,
0,
BUS_DMA_WAITOK|BUS_DMA_ALLOCNOW,
@@ -264,7 +273,7 @@ ld_virtio_attach(device_t parent, device_t self, void *aux)
struct ld_softc *ld = &sc->sc_ld;
struct virtio_softc *vsc = device_private(parent);
uint64_t features;
- int qsize, maxxfersize, maxnsegs;
+ int qsize, maxxfersize;
if (virtio_child(vsc) != NULL) {
aprint_normal(": child already attached for %s; "
@@ -317,30 +326,33 @@ ld_virtio_attach(device_t parent, device_t self, void *aux)
maxxfersize = MAXPHYS;
if (features & VIRTIO_BLK_F_SEG_MAX) {
- maxnsegs = virtio_read_device_config_4(vsc,
+ ld->sc_maxnsegs = virtio_read_device_config_4(vsc,
VIRTIO_BLK_CONFIG_SEG_MAX);
- if (maxnsegs < VIRTIO_BLK_MIN_SEGMENTS) {
+ if (ld->sc_maxnsegs < VIRTIO_BLK_MIN_SEGMENTS) {
aprint_error_dev(sc->sc_dev,
"Too small SEG_MAX %d minimum is %d\n",
- maxnsegs, VIRTIO_BLK_MIN_SEGMENTS);
- maxnsegs = maxxfersize / NBPG;
+ ld->sc_maxnsegs, VIRTIO_BLK_MIN_SEGMENTS);
+ ld->sc_maxnsegs = maxxfersize / NBPG;
// goto err;
}
} else
- maxnsegs = maxxfersize / NBPG;
+ /*
+ * if there is no VIRTIO_BLK_F_SEG_MAX feature advertised, the
+ * number of segments can be as low as 1 (i.e. Firecracker)
+ */
+ ld->sc_maxnsegs = 1;
/* 2 for the minimum size */
- maxnsegs += VIRTIO_BLK_MIN_SEGMENTS;
+ ld->sc_maxnsegs += VIRTIO_BLK_MIN_SEGMENTS;
virtio_init_vq_vqdone(vsc, &sc->sc_vq, 0,
ld_virtio_vq_done);
- if (virtio_alloc_vq(vsc, &sc->sc_vq, maxxfersize, maxnsegs,
+ if (virtio_alloc_vq(vsc, &sc->sc_vq, maxxfersize, ld->sc_maxnsegs,
"I/O request") != 0) {
goto err;
}
qsize = sc->sc_vq.vq_num;
-
if (virtio_child_attach_finish(vsc, &sc->sc_vq, 1,
NULL, VIRTIO_F_INTR_MSIX) != 0)
goto err;
diff --git a/sys/dev/pv/files.pv b/sys/dev/pv/files.pv
new file mode 100644
index 00000000000..1775b9edf5a
--- /dev/null
+++ b/sys/dev/pv/files.pv
@@ -0,0 +1,8 @@
+define pvbus {}
+
+device pv {}
+attach pv at pvbus
+file dev/pv/pvbus.c pvbus needs-flag
+
+attach virtio at pv with mmio_cmdline: virtio_mmio
+file dev/virtio/virtio_mmio_cmdline.c mmio_cmdline
diff --git a/sys/dev/pv/pvbus.c b/sys/dev/pv/pvbus.c
new file mode 100644
index 00000000000..f1101d62305
--- /dev/null
+++ b/sys/dev/pv/pvbus.c
@@ -0,0 +1,71 @@
+/* $NetBSD$ */
+
+/*-
+ * Copyright (c) 2024 The NetBSD Foundation, Inc.
+ * All rights reserved.
+ *
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Emile 'iMil' Heitor.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <sys/cdefs.h>
+#include <sys/param.h>
+#include <sys/bus.h>
+#include <sys/device.h>
+#include <sys/kernel.h>
+#include <sys/systm.h>
+
+#include <machine/bus_private.h>
+
+#include <dev/pv/pvvar.h>
+
+struct x86_bus_dma_tag pvbus_bus_dma_tag = {
+ ._tag_needs_free = 0,
+ ._bounce_thresh = 0,
+ ._bounce_alloc_lo = 0,
+ ._bounce_alloc_hi = 0,
+ ._may_bounce = NULL,
+};
+
+static int
+pv_match(device_t parent, cfdata_t match, void *aux)
+{
+ return 1;
+}
+
+static void
+pv_attach(device_t parent, device_t self, void *aux) {
+ struct pv_attach_args pvaa;
+
+ pvaa.pvaa_memt = x86_bus_space_mem;
+ pvaa.pvaa_dmat = &pvbus_bus_dma_tag;
+
+ aprint_naive("\n");
+ aprint_normal("\n");
+
+ config_found(self, &pvaa, NULL, CFARGS_NONE);
+}
+
+CFATTACH_DECL_NEW(pv, sizeof(struct pv_softc),
+ pv_match, pv_attach, NULL, NULL);
diff --git a/sys/dev/pv/pvvar.h b/sys/dev/pv/pvvar.h
new file mode 100644
index 00000000000..bba44b926dc
--- /dev/null
+++ b/sys/dev/pv/pvvar.h
@@ -0,0 +1,48 @@
+/* $NetBSD$ */
+
+/*-
+ * Copyright (c) 2024 The NetBSD Foundation, Inc.
+ * All rights reserved.
+ *
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Emile 'iMil' Heitor.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _PVBUS_PVVAR_H_
+#define _PVBUS_PVVAR_H_
+
+struct pv_softc {
+ device_t sc_dev;
+};
+
+struct pvbus_attach_args {
+ const char *pvba_busname;
+};
+
+struct pv_attach_args {
+ bus_space_tag_t pvaa_memt;
+ bus_dma_tag_t pvaa_dmat;
+};
+
+#endif
diff --git a/sys/dev/virtio/virtio_mmio_cmdline.c b/sys/dev/virtio/virtio_mmio_cmdline.c
new file mode 100644
index 00000000000..ec2fadc8066
--- /dev/null
+++ b/sys/dev/virtio/virtio_mmio_cmdline.c
@@ -0,0 +1,345 @@
+/* $NetBSD$ */
+
+/*-
+ * Copyright (c) 2024 The NetBSD Foundation, Inc.
+ * All rights reserved.
+ *
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Emile 'iMil' Heitor.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
+ * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
+ * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
+ * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/*-
+ * Copyright (c) 2022 Colin Percival
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <sys/param.h>
+#include <sys/bus.h>
+#include <sys/device.h>
+#include <sys/kernel.h>
+#include <sys/module.h>
+#include <sys/systm.h>
+
+#define VIRTIO_PRIVATE
+#include <dev/virtio/virtio_mmiovar.h>
+#include <dev/pv/pvvar.h>
+#include <xen/hypervisor.h>
+
+#include <machine/i82093var.h>
+#include "ioapic.h"
+
+#define VMMIOSTR "virtio_mmio.device="
+
+struct mmio_args {
+ uint64_t sz;
+ uint64_t baseaddr;
+ uint64_t irq;
+ uint64_t id;
+};
+
+struct virtio_mmio_cmdline_softc {
+ struct virtio_mmio_softc sc_msc;
+ struct mmio_args margs;
+};
+
+static int virtio_mmio_cmdline_match(device_t, cfdata_t, void *);
+static void virtio_mmio_cmdline_attach(device_t, device_t, void *);
+static int virtio_mmio_cmdline_do_attach(device_t,
+ struct pv_attach_args *,
+ struct mmio_args *);
+static int virtio_mmio_cmdline_detach(device_t, int);
+static int virtio_mmio_cmdline_rescan(device_t, const char *, const int *);
+static int virtio_mmio_cmdline_alloc_interrupts(struct virtio_mmio_softc *);
+static void virtio_mmio_cmdline_free_interrupts(struct virtio_mmio_softc *);
+
+CFATTACH_DECL3_NEW(mmio_cmdline,
+ sizeof(struct virtio_mmio_cmdline_softc),
+ virtio_mmio_cmdline_match, virtio_mmio_cmdline_attach,
+ virtio_mmio_cmdline_detach, NULL,
+ virtio_mmio_cmdline_rescan, (void *)voidop, DVF_DETACH_SHUTDOWN);
+
+static int
+virtio_mmio_cmdline_match(device_t parent, cfdata_t match, void *aux)
+{
+ if (strstr(xen_start_info.cmd_line, VMMIOSTR) == NULL)
+ return 0;
+
+ return 1;
+}
+
+static void
+parsearg(struct mmio_args *margs, const char *arg)
+{
+ char *p;
+
+ /* <size> */
+ margs->sz = strtoull(arg, (char **)&p, 0);
+ if ((margs->sz == 0) || (margs->sz == ULLONG_MAX))
+ goto bad;
+ switch (*p) {
+ case 'E': case 'e':
+ margs->sz <<= 10;
+ /* FALLTHROUGH */
+ case 'P': case 'p':
+ margs->sz <<= 10;
+ /* FALLTHROUGH */
+ case 'T': case 't':
+ margs->sz <<= 10;
+ /* FALLTHROUGH */
+ case 'G': case 'g':
+ margs->sz <<= 10;
+ /* FALLTHROUGH */
+ case 'M': case 'm':
+ margs->sz <<= 10;
+ /* FALLTHROUGH */
+ case 'K': case 'k':
+ margs->sz <<= 10;
+ p++;
+ break;
+ }
+
+ /* @<baseaddr> */
+ if (*p++ != '@')
+ goto bad;
+ margs->baseaddr = strtoull(p, (char **)&p, 0);
+ if ((margs->baseaddr == 0) || (margs->baseaddr == ULLONG_MAX))
+ goto bad;
+
+ /* :<irq> */
+ if (*p++ != ':')
+ goto bad;
+ margs->irq = strtoull(p, (char **)&p, 0);
+ if ((margs->irq == 0) || (margs->irq == ULLONG_MAX))
+ goto bad;
+
+ /* Optionally, :<id> */
+ if (*p) {
+ if (*p++ != ':')
+ goto bad;
+ margs->id = strtoull(p, (char **)&p, 0);
+ if ((margs->id == 0) || (margs->id == ULLONG_MAX))
+ goto bad;
+ } else {
+ margs->id = 0;
+ }
+
+ /* Should have reached the end of the string. */
+ if (*p)
+ goto bad;
+
+ return;
+
+bad:
+ aprint_error("Error parsing virtio_mmio parameter: %s\n", arg);
+}
+
+static void
+virtio_mmio_cmdline_attach(device_t parent, device_t self, void *aux)
+{
+ struct virtio_mmio_cmdline_softc *sc = device_private(self);
+ struct pv_attach_args *pvaa = aux;
+ struct mmio_args *margs = &sc->margs;
+ char *v, *n, cmdline[128];
+ int error;
+ static char *p = NULL;
+ static int idx = 0;
+ bool hasnext;
+
+ aprint_normal("\n");
+ aprint_naive("\n");
+
+ if (idx == 0) {
+ strncpy(cmdline, xen_start_info.cmd_line, sizeof(cmdline));
+ aprint_verbose("kernel parameters: %s\n", cmdline);
+ if ((p = strstr(cmdline, VMMIOSTR)) == NULL)
+ return;
+ }
+
+ while (*p) { /* manual strtok() */
+ hasnext = false;
+ v = p; /* start of VMMIOSTR or next param */
+ while (*p && *p != ' ') /* find end of param */
+ p++;
+ if (*p) {
+ n = p; /* record end of param */
+ *p = '\0'; /* end it */
+ hasnext = true; /* more params to come */
+ }
+ if (strncmp(v, VMMIOSTR, strlen(VMMIOSTR)) != 0)
+ continue; /* this was not an MMIO parameter */
+ p = v; /* start of VMMIOSTR */
+ while (*p && *p != '=') /* point to the value */
+ p++;
+ if (*p) {
+ p++;
+ aprint_normal("viommio: %s\n", p);
+ parsearg(margs, p);
+
+ error = virtio_mmio_cmdline_do_attach(self,
+ pvaa, margs);
+
+ if (error)
+ return;
+ }
+ if (hasnext) {
+ p = n+1; /* previously recorded end of param */
+ idx++;
+ config_found(parent, pvaa, NULL, CFARGS_NONE);
+ }
+ }
+}
+
+static int
+virtio_mmio_cmdline_do_attach(device_t self,
+ struct pv_attach_args *pvaa,
+ struct mmio_args *margs)
+{
+ struct virtio_mmio_cmdline_softc *sc = device_private(self);
+ struct virtio_mmio_softc *const msc = &sc->sc_msc;
+ struct virtio_softc *const vsc = &msc->sc_sc;
+ int error;
+
+ msc->sc_iot = pvaa->pvaa_memt;
+ vsc->sc_dmat = pvaa->pvaa_dmat;
+ msc->sc_iosize = margs->sz;
+ vsc->sc_dev = self;
+
+ error = bus_space_map(
+ msc->sc_iot, margs->baseaddr,
+ margs->sz, 0, &msc->sc_ioh
+ );
+ if (error) {
+ aprint_error_dev(self,
+ "couldn't map %#" PRIx64 ": %d",
+ (uint64_t)margs->baseaddr, error
+ );
+ return error;
+ }
+
+ msc->sc_alloc_interrupts = virtio_mmio_cmdline_alloc_interrupts;
+ msc->sc_free_interrupts = virtio_mmio_cmdline_free_interrupts;
+
+ virtio_mmio_common_attach(msc);
+ virtio_mmio_cmdline_rescan(self, "virtio", NULL);
+
+ return 0;
+}
+
+static int
+virtio_mmio_cmdline_detach(device_t self, int flags)
+{
+ struct virtio_mmio_cmdline_softc * const sc = device_private(self);
+ struct virtio_mmio_softc * const msc = &sc->sc_msc;
+
+ return virtio_mmio_common_detach(msc, flags);
+}
+
+static int
+virtio_mmio_cmdline_rescan(device_t self, const char *ifattr, const int *locs)
+{
+ struct virtio_mmio_cmdline_softc *const sc = device_private(self);
+ struct virtio_mmio_softc *const msc = &sc->sc_msc;
+ struct virtio_softc *const vsc = &msc->sc_sc;
+ struct virtio_attach_args va;
+
+ if (vsc->sc_child)
+ return 0;
+
+ memset(&va, 0, sizeof(va));
+ va.sc_childdevid = vsc->sc_childdevid;
+
+ config_found(self, &va, NULL, CFARGS_NONE);
+
+ if (virtio_attach_failed(vsc))
+ return 0;
+
+ return 0;
+}
+
+
+static int
+virtio_mmio_cmdline_alloc_interrupts(struct virtio_mmio_softc *msc)
+{
+ struct virtio_mmio_cmdline_softc *const sc =
+ (struct virtio_mmio_cmdline_softc *)msc;
+ struct virtio_softc *const vsc = &msc->sc_sc;
+ struct ioapic_softc *ioapic;
+ struct pic *pic;
+ int irq = sc->margs.irq;
+ int pin = irq;
+ bool mpsafe;
+
+ ioapic = ioapic_find_bybase(irq);
+
+ if (ioapic != NULL) {
+ KASSERT(ioapic->sc_pic.pic_type == PIC_IOAPIC);
+ pic = &ioapic->sc_pic;
+ pin = irq - pic->pic_vecbase;
+ irq = -1;
+ } else
+ pic = &i8259_pic;
+
+ mpsafe = (0 != (vsc->sc_flags & VIRTIO_F_INTR_MPSAFE));
+
+ msc->sc_ih = intr_establish_xname(irq, pic, pin, IST_LEVEL, vsc->sc_ipl,
+ virtio_mmio_intr, msc, mpsafe, device_xname(vsc->sc_dev));
+ if (msc->sc_ih == NULL) {
+ aprint_error_dev(vsc->sc_dev,
+ "failed to establish interrupt\n");
+ return -1;
+ }
+ aprint_normal_dev(vsc->sc_dev, "interrupting on %d\n", irq);
+
+ return 0;
+}
+
+static void
+virtio_mmio_cmdline_free_interrupts(struct virtio_mmio_softc *msc)
+{
+ if (msc->sc_ih != NULL) {
+ intr_disestablish(msc->sc_ih);
+ msc->sc_ih = NULL;
+ }
+}
+