Index: user/marcel/libvdsk/bhyve/bhyve.8 =================================================================== --- user/marcel/libvdsk/bhyve/bhyve.8 (revision 279574) +++ user/marcel/libvdsk/bhyve/bhyve.8 (revision 279575) @@ -1,322 +1,324 @@ .\" Copyright (c) 2013 Peter Grehan .\" 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 AUTHORS 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 AUTHORS 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. .\" .\" $FreeBSD$ .\" .Dd September 17, 2014 .Dt BHYVE 8 .Os .Sh NAME .Nm bhyve .Nd "run a guest operating system inside a virtual machine" .Sh SYNOPSIS .Nm -.Op Fl abehwxACHPWY +.Op Fl abehuwxACHPWY .Op Fl c Ar numcpus .Op Fl g Ar gdbport .Op Fl l Ar lpcdev Ns Op , Ns Ar conf .Op Fl m Ar size Ns Op Ar K|k|M|m|G|g|T|t .Op Fl p Ar vcpu:hostcpu .Op Fl s Ar slot,emulation Ns Op , Ns Ar conf .Op Fl U Ar uuid .Ar vmname .Sh DESCRIPTION .Nm is a hypervisor that runs guest operating systems inside a virtual machine. .Pp Parameters such as the number of virtual CPUs, amount of guest memory, and I/O connectivity can be specified with command-line parameters. .Pp The guest operating system must be loaded with .Xr bhyveload 4 or a similar boot loader before running .Nm . .Pp .Nm runs until the guest operating system reboots or an unhandled hypervisor exit is detected. .Sh OPTIONS .Bl -tag -width 10n .It Fl a The guest's local APIC is configured in xAPIC mode. The xAPIC mode is the default setting so this option is redundant. It will be deprecated in a future version. .It Fl A Generate ACPI tables. Required for .Fx Ns /amd64 guests. .It Fl b Enable a low-level console device supported by .Fx kernels compiled with .Cd "device bvmconsole" . This option will be deprecated in a future version. .It Fl c Ar numcpus Number of guest virtual CPUs. The default is 1 and the maximum is 16. .It Fl C Include guest memory in core file. .It Fl e Force .Nm to exit when a guest issues an access to an I/O port that is not emulated. This is intended for debug purposes. .It Fl g Ar gdbport For .Fx kernels compiled with .Cd "device bvmdebug" , allow a remote kernel kgdb to be relayed to the guest kernel gdb stub via a local IPv4 address and this port. This option will be deprecated in a future version. .It Fl h Print help message and exit. .It Fl H Yield the virtual CPU thread when a HLT instruction is detected. If this option is not specified, virtual CPUs will use 100% of a host CPU. .It Fl l Ar lpcdev Ns Op , Ns Ar conf Allow devices behind the LPC PCI-ISA bridge to be configured. The only supported devices are the TTY-class devices, .Li com1 and .Li com2 . .It Fl m Ar size Ns Op Ar K|k|M|m|G|g|T|t Guest physical memory size in bytes. This must be the same size that was given to .Xr bhyveload 8 . .Pp The size argument may be suffixed with one of K, M, G or T (either upper or lower case) to indicate a multiple of kilobytes, megabytes, gigabytes, or terabytes. If no suffix is given, the value is assumed to be in megabytes. .It Fl p Ar vcpu:hostcpu Pin guest's virtual CPU .Em vcpu to .Em hostcpu . .It Fl P Force the guest virtual CPU to exit when a PAUSE instruction is detected. .It Fl s Ar slot,emulation Ns Op , Ns Ar conf Configure a virtual PCI slot and function. .Pp .Nm bhyve provides PCI bus emulation and virtual devices that can be attached to slots on the bus. There are 32 available slots, with the option of providing up to 8 functions per slot. .Bl -tag -width 10n .It Ar slot .Ar pcislot[:function] .Ar bus:pcislot:function .Pp The .Ar pcislot value is 0 to 31. The optional function value is 0 to 7. The optional .Ar bus value is 0 to 255. If not specified, the function value defaults to 0. If not specified, the bus value defaults to 0. .It Ar emulation .Bl -tag -width 10n .It Li hostbridge | Li amd_hostbridge .Pp Provide a simple host bridge. This is usually configured at slot 0, and is required by most guest operating systems. The .Li amd_hostbridge emulation is identical but uses a PCI vendor ID of .Li AMD . .It Li passthru PCI pass-through device. .It Li virtio-net Virtio network interface. .It Li virtio-blk Virtio block storage interface. .It Li virtio-rnd Virtio RNG interface. .It Li ahci-cd AHCI controller attached to an ATAPI CD/DVD. .It Li ahci-hd AHCI controller attached to a SATA hard-drive. .It Li uart PCI 16550 serial device. .It Li lpc LPC PCI-ISA bridge with COM1 and COM2 16550 serial ports. The LPC bridge emulation can only be configured on bus 0. .El .It Op Ar conf This optional parameter describes the backend for device emulations. If .Ar conf is not specified, the device emulation has no backend and can be considered unconnected. .Pp Network devices: .Bl -tag -width 10n .It Ar tapN Ns Op , Ns Ar mac=xx:xx:xx:xx:xx:xx .It Ar vmnetN Ns Op , Ns Ar mac=xx:xx:xx:xx:xx:xx .Pp If .Ar mac is not specified, the MAC address is derived from a fixed OUI and the remaining bytes from an MD5 hash of the slot and function numbers and the device name. .Pp The MAC address is an ASCII string in .Xr ethers 5 format. .El .Pp Block storage devices: .Bl -tag -width 10n .It Pa /filename Ns Oo , Ns Li nocache Oc Ns Oo , Ns Li direct Oc Ns Oo , Ns Li ro Oc .It Pa /dev/xxx Ns Oo , Ns Ar nocache Oc Ns Oo , Ns Ar direct Oc Ns Oo , Ns Ar ro Oc .Bl -tag -width 8n .It Li nocache Open the file with .Dv O_DIRECT . .It Li direct Open the file using .Dv O_SYNC . .It Li ro Force the file to be opened read-only. .El .Pp The .Li nocache , .Li direct , and .Li ro options are not available for virtio block devices. .El .Pp TTY devices: .Bl -tag -width 10n .It Li stdio Connect the serial port to the standard input and output of the bhyve process. .It Pa /dev/xxx Use the host TTY device for serial port I/O. .El .Pp Pass-through devices: .Bl -tag -width 10n .It Ns Ar slot Ns / Ns Ar bus Ns / Ns Ar function Connect to a PCI device on the host at the selector described by .Ar slot , .Ar bus , and .Ar function numbers. .El .Pp The host device must have been reserved at boot-time using the .Va pptdev loader variable as described in .Xr vmm 4 . .El +.It Fl u +RTC keeps UTC time. .It Fl U Ar uuid Set the universally unique identifier .Pq UUID in the guest's System Management BIOS System Information structure. By default a UUID is generated from the host's hostname and .Ar vmname . .It Fl w Ignore accesses to unimplemented Model Specific Registers (MSRs). This is intended for debug purposes. .It Fl W Force virtio PCI device emulations to use MSI interrupts instead of MSI-X interrupts. .It Fl x The guest's local APIC is configured in x2APIC mode. .It Fl Y Disable MPtable generation. .It Ar vmname Alphanumeric name of the guest. This should be the same as that created by .Xr bhyveload 8 . .El .Sh EXAMPLES The guest operating system must have been loaded with .Xr bhyveload 4 or a similar boot loader before .Xr bhyve 4 can be run. .Pp To run a virtual machine with 1GB of memory, two virtual CPUs, a virtio block device backed by the .Pa /my/image filesystem image, and a serial port for the console: .Bd -literal -offset indent bhyve -c 2 -s 0,hostbridge -s 1,lpc -s 2,virtio-blk,/my/image \\ -l com1,stdio -A -H -P -m 1G vm1 .Ed .Pp Run a 24GB single-CPU virtual machine with three network ports, one of which has a MAC address specified: .Bd -literal -offset indent bhyve -s 0,hostbridge -s 1,lpc -s 2:0,virtio-net,tap0 \\ -s 2:1,virtio-net,tap1 \\ -s 2:2,virtio-net,tap2,mac=00:be:fa:76:45:00 \\ -s 3,virtio-blk,/my/image -l com1,stdio \\ -A -H -P -m 24G bigvm .Ed .Pp Run an 8GB quad-CPU virtual machine with 8 AHCI SATA disks, an AHCI ATAPI CD-ROM, a single virtio network port, an AMD hostbridge, and the console port connected to an .Xr nmdm 4 null-model device. .Bd -literal -offset indent bhyve -c 4 \e\ -s 0,amd_hostbridge -s 1,lpc \\ -s 1:0,ahci-hd,/images/disk.1 \\ -s 1:1,ahci-hd,/images/disk.2 \\ -s 1:2,ahci-hd,/images/disk.3 \\ -s 1:3,ahci-hd,/images/disk.4 \\ -s 1:4,ahci-hd,/images/disk.5 \\ -s 1:5,ahci-hd,/images/disk.6 \\ -s 1:6,ahci-hd,/images/disk.7 \\ -s 1:7,ahci-hd,/images/disk.8 \\ -s 2,ahci-cd,/images.install.iso \\ -s 3,virtio-net,tap0 \\ -l com1,/dev/nmdm0A \\ -A -H -P -m 8G .Ed .Sh SEE ALSO .Xr bhyve 4 , .Xr nmdm 4 , .Xr vmm 4 , .Xr ethers 5 , .Xr bhyvectl 8 , .Xr bhyveload 8 .Sh HISTORY .Nm first appeared in .Fx 10.0 . .Sh AUTHORS .An Neel Natu Aq Mt neel@freebsd.org .An Peter Grehan Aq Mt grehan@freebsd.org Index: user/marcel/libvdsk/bhyve/bhyverun.c =================================================================== --- user/marcel/libvdsk/bhyve/bhyverun.c (revision 279574) +++ user/marcel/libvdsk/bhyve/bhyverun.c (revision 279575) @@ -1,883 +1,880 @@ /*- * Copyright (c) 2011 NetApp, 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 NETAPP, INC ``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 NETAPP, INC 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. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "bhyverun.h" #include "acpi.h" #include "inout.h" #include "dbgport.h" #include "ioapic.h" #include "mem.h" #include "mevent.h" #include "mptbl.h" #include "pci_emul.h" #include "pci_irq.h" #include "pci_lpc.h" #include "smbiostbl.h" #include "xmsr.h" #include "spinup_ap.h" #include "rtc.h" #define GUEST_NIO_PORT 0x488 /* guest upcalls via i/o port */ #define MB (1024UL * 1024) #define GB (1024UL * MB) typedef int (*vmexit_handler_t)(struct vmctx *, struct vm_exit *, int *vcpu); extern int vmexit_task_switch(struct vmctx *, struct vm_exit *, int *vcpu); char *vmname; int guest_ncpus; char *guest_uuid_str; static int guest_vmexit_on_hlt, guest_vmexit_on_pause; static int virtio_msix = 1; static int x2apic_mode = 0; /* default is xAPIC */ static int strictio; static int strictmsr = 1; static int acpi; static char *progname; static const int BSP = 0; static cpuset_t cpumask; static void vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip); static struct vm_exit vmexit[VM_MAXCPU]; struct bhyvestats { uint64_t vmexit_bogus; uint64_t vmexit_bogus_switch; uint64_t vmexit_hlt; uint64_t vmexit_pause; uint64_t vmexit_mtrap; uint64_t vmexit_inst_emul; uint64_t cpu_switch_rotate; uint64_t cpu_switch_direct; } stats; struct mt_vmm_info { pthread_t mt_thr; struct vmctx *mt_ctx; int mt_vcpu; } mt_vmm_info[VM_MAXCPU]; static cpuset_t *vcpumap[VM_MAXCPU] = { NULL }; static void usage(int code) { fprintf(stderr, - "Usage: %s [-abehwxACHPWY] [-c vcpus] [-g ] [-l ]\n" + "Usage: %s [-abehuwxACHPWY] [-c vcpus] [-g ] [-l ]\n" " %*s [-m mem] [-p vcpu:hostcpu] [-s ] [-U uuid] \n" " -a: local apic is in xAPIC mode (deprecated)\n" " -A: create ACPI tables\n" " -c: # cpus (default 1)\n" " -C: include guest memory in core file\n" " -e: exit on unhandled I/O access\n" " -g: gdb port\n" " -h: help\n" " -H: vmexit from the guest on hlt\n" " -l: LPC device configuration\n" " -m: memory size in MB\n" " -p: pin 'vcpu' to 'hostcpu'\n" " -P: vmexit from the guest on pause\n" " -s: PCI slot config\n" + " -u: RTC keeps UTC time\n" " -U: uuid\n" " -w: ignore unimplemented MSRs\n" " -W: force virtio to use single-vector MSI\n" " -x: local apic is in x2APIC mode\n" " -Y: disable MPtable generation\n", progname, (int)strlen(progname), ""); exit(code); } static int pincpu_parse(const char *opt) { int vcpu, pcpu; if (sscanf(opt, "%d:%d", &vcpu, &pcpu) != 2) { fprintf(stderr, "invalid format: %s\n", opt); return (-1); } if (vcpu < 0 || vcpu >= VM_MAXCPU) { fprintf(stderr, "vcpu '%d' outside valid range from 0 to %d\n", vcpu, VM_MAXCPU - 1); return (-1); } if (pcpu < 0 || pcpu >= CPU_SETSIZE) { fprintf(stderr, "hostcpu '%d' outside valid range from " "0 to %d\n", pcpu, CPU_SETSIZE - 1); return (-1); } if (vcpumap[vcpu] == NULL) { if ((vcpumap[vcpu] = malloc(sizeof(cpuset_t))) == NULL) { perror("malloc"); return (-1); } CPU_ZERO(vcpumap[vcpu]); } CPU_SET(pcpu, vcpumap[vcpu]); return (0); } void vm_inject_fault(void *arg, int vcpu, int vector, int errcode_valid, int errcode) { struct vmctx *ctx; - int error; + int error, restart_instruction; ctx = arg; - if (errcode_valid) - error = vm_inject_exception2(ctx, vcpu, vector, errcode); - else - error = vm_inject_exception(ctx, vcpu, vector); - assert(error == 0); + restart_instruction = 1; - /* - * Set the instruction length to 0 to ensure that the instruction is - * restarted when the fault handler returns. - */ - vmexit[vcpu].inst_length = 0; + error = vm_inject_exception(ctx, vcpu, vector, errcode_valid, errcode, + restart_instruction); + assert(error == 0); } void * paddr_guest2host(struct vmctx *ctx, uintptr_t gaddr, size_t len) { return (vm_map_gpa(ctx, gaddr, len)); } int fbsdrun_vmexit_on_pause(void) { return (guest_vmexit_on_pause); } int fbsdrun_vmexit_on_hlt(void) { return (guest_vmexit_on_hlt); } int fbsdrun_virtio_msix(void) { return (virtio_msix); } static void * fbsdrun_start_thread(void *param) { char tname[MAXCOMLEN + 1]; struct mt_vmm_info *mtp; int vcpu; mtp = param; vcpu = mtp->mt_vcpu; snprintf(tname, sizeof(tname), "vcpu %d", vcpu); pthread_set_name_np(mtp->mt_thr, tname); vm_loop(mtp->mt_ctx, vcpu, vmexit[vcpu].rip); /* not reached */ exit(1); return (NULL); } void fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip) { int error; assert(fromcpu == BSP); /* * The 'newcpu' must be activated in the context of 'fromcpu'. If * vm_activate_cpu() is delayed until newcpu's pthread starts running * then vmm.ko is out-of-sync with bhyve and this can create a race * with vm_suspend(). */ error = vm_activate_cpu(ctx, newcpu); assert(error == 0); CPU_SET_ATOMIC(newcpu, &cpumask); /* * Set up the vmexit struct to allow execution to start * at the given RIP */ vmexit[newcpu].rip = rip; vmexit[newcpu].inst_length = 0; mt_vmm_info[newcpu].mt_ctx = ctx; mt_vmm_info[newcpu].mt_vcpu = newcpu; error = pthread_create(&mt_vmm_info[newcpu].mt_thr, NULL, fbsdrun_start_thread, &mt_vmm_info[newcpu]); assert(error == 0); } static int fbsdrun_deletecpu(struct vmctx *ctx, int vcpu) { if (!CPU_ISSET(vcpu, &cpumask)) { fprintf(stderr, "Attempting to delete unknown cpu %d\n", vcpu); exit(1); } CPU_CLR_ATOMIC(vcpu, &cpumask); return (CPU_EMPTY(&cpumask)); } static int vmexit_handle_notify(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu, uint32_t eax) { #if BHYVE_DEBUG /* * put guest-driven debug here */ #endif return (VMEXIT_CONTINUE); } static int vmexit_inout(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) { int error; int bytes, port, in, out, string; int vcpu; vcpu = *pvcpu; port = vme->u.inout.port; bytes = vme->u.inout.bytes; string = vme->u.inout.string; in = vme->u.inout.in; out = !in; /* Extra-special case of host notifications */ if (out && port == GUEST_NIO_PORT) { error = vmexit_handle_notify(ctx, vme, pvcpu, vme->u.inout.eax); return (error); } error = emulate_inout(ctx, vcpu, vme, strictio); - if (!error && in && !string) { - error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, - vme->u.inout.eax); - assert(error == 0); - } - if (error) { fprintf(stderr, "Unhandled %s%c 0x%04x\n", in ? "in" : "out", bytes == 1 ? 'b' : (bytes == 2 ? 'w' : 'l'), port); return (VMEXIT_ABORT); } else { return (VMEXIT_CONTINUE); } } static int vmexit_rdmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) { uint64_t val; uint32_t eax, edx; int error; val = 0; error = emulate_rdmsr(ctx, *pvcpu, vme->u.msr.code, &val); if (error != 0) { fprintf(stderr, "rdmsr to register %#x on vcpu %d\n", vme->u.msr.code, *pvcpu); if (strictmsr) { vm_inject_gp(ctx, *pvcpu); - return (VMEXIT_RESTART); + return (VMEXIT_CONTINUE); } } eax = val; error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RAX, eax); assert(error == 0); edx = val >> 32; error = vm_set_register(ctx, *pvcpu, VM_REG_GUEST_RDX, edx); assert(error == 0); return (VMEXIT_CONTINUE); } static int vmexit_wrmsr(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) { int error; error = emulate_wrmsr(ctx, *pvcpu, vme->u.msr.code, vme->u.msr.wval); if (error != 0) { fprintf(stderr, "wrmsr to register %#x(%#lx) on vcpu %d\n", vme->u.msr.code, vme->u.msr.wval, *pvcpu); if (strictmsr) { vm_inject_gp(ctx, *pvcpu); - return (VMEXIT_RESTART); + return (VMEXIT_CONTINUE); } } return (VMEXIT_CONTINUE); } static int vmexit_spinup_ap(struct vmctx *ctx, struct vm_exit *vme, int *pvcpu) { int newcpu; int retval = VMEXIT_CONTINUE; newcpu = spinup_ap(ctx, *pvcpu, vme->u.spinup_ap.vcpu, vme->u.spinup_ap.rip); return (retval); } #define DEBUG_EPT_MISCONFIG #ifdef DEBUG_EPT_MISCONFIG #define EXIT_REASON_EPT_MISCONFIG 49 #define VMCS_GUEST_PHYSICAL_ADDRESS 0x00002400 #define VMCS_IDENT(x) ((x) | 0x80000000) static uint64_t ept_misconfig_gpa, ept_misconfig_pte[4]; static int ept_misconfig_ptenum; #endif static int vmexit_vmx(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { fprintf(stderr, "vm exit[%d]\n", *pvcpu); fprintf(stderr, "\treason\t\tVMX\n"); fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); fprintf(stderr, "\tstatus\t\t%d\n", vmexit->u.vmx.status); fprintf(stderr, "\texit_reason\t%u\n", vmexit->u.vmx.exit_reason); fprintf(stderr, "\tqualification\t0x%016lx\n", vmexit->u.vmx.exit_qualification); fprintf(stderr, "\tinst_type\t\t%d\n", vmexit->u.vmx.inst_type); fprintf(stderr, "\tinst_error\t\t%d\n", vmexit->u.vmx.inst_error); #ifdef DEBUG_EPT_MISCONFIG if (vmexit->u.vmx.exit_reason == EXIT_REASON_EPT_MISCONFIG) { vm_get_register(ctx, *pvcpu, VMCS_IDENT(VMCS_GUEST_PHYSICAL_ADDRESS), &ept_misconfig_gpa); vm_get_gpa_pmap(ctx, ept_misconfig_gpa, ept_misconfig_pte, &ept_misconfig_ptenum); fprintf(stderr, "\tEPT misconfiguration:\n"); fprintf(stderr, "\t\tGPA: %#lx\n", ept_misconfig_gpa); fprintf(stderr, "\t\tPTE(%d): %#lx %#lx %#lx %#lx\n", ept_misconfig_ptenum, ept_misconfig_pte[0], ept_misconfig_pte[1], ept_misconfig_pte[2], ept_misconfig_pte[3]); } #endif /* DEBUG_EPT_MISCONFIG */ return (VMEXIT_ABORT); } static int vmexit_svm(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { fprintf(stderr, "vm exit[%d]\n", *pvcpu); fprintf(stderr, "\treason\t\tSVM\n"); fprintf(stderr, "\trip\t\t0x%016lx\n", vmexit->rip); fprintf(stderr, "\tinst_length\t%d\n", vmexit->inst_length); fprintf(stderr, "\texitcode\t%#lx\n", vmexit->u.svm.exitcode); fprintf(stderr, "\texitinfo1\t%#lx\n", vmexit->u.svm.exitinfo1); fprintf(stderr, "\texitinfo2\t%#lx\n", vmexit->u.svm.exitinfo2); return (VMEXIT_ABORT); } static int vmexit_bogus(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { + assert(vmexit->inst_length == 0); + stats.vmexit_bogus++; - return (VMEXIT_RESTART); + return (VMEXIT_CONTINUE); } static int vmexit_hlt(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { stats.vmexit_hlt++; /* * Just continue execution with the next instruction. We use * the HLT VM exit as a way to be friendly with the host * scheduler. */ return (VMEXIT_CONTINUE); } static int vmexit_pause(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { stats.vmexit_pause++; return (VMEXIT_CONTINUE); } static int vmexit_mtrap(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { + assert(vmexit->inst_length == 0); + stats.vmexit_mtrap++; - return (VMEXIT_RESTART); + return (VMEXIT_CONTINUE); } static int vmexit_inst_emul(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { int err; stats.vmexit_inst_emul++; err = emulate_mem(ctx, *pvcpu, vmexit->u.inst_emul.gpa, &vmexit->u.inst_emul.vie, &vmexit->u.inst_emul.paging); if (err) { if (err == EINVAL) { fprintf(stderr, "Failed to emulate instruction at 0x%lx\n", vmexit->rip); } else if (err == ESRCH) { fprintf(stderr, "Unhandled memory access to 0x%lx\n", vmexit->u.inst_emul.gpa); } return (VMEXIT_ABORT); } return (VMEXIT_CONTINUE); } static pthread_mutex_t resetcpu_mtx = PTHREAD_MUTEX_INITIALIZER; static pthread_cond_t resetcpu_cond = PTHREAD_COND_INITIALIZER; static int vmexit_suspend(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { enum vm_suspend_how how; how = vmexit->u.suspended.how; fbsdrun_deletecpu(ctx, *pvcpu); if (*pvcpu != BSP) { pthread_mutex_lock(&resetcpu_mtx); pthread_cond_signal(&resetcpu_cond); pthread_mutex_unlock(&resetcpu_mtx); pthread_exit(NULL); } pthread_mutex_lock(&resetcpu_mtx); while (!CPU_EMPTY(&cpumask)) { pthread_cond_wait(&resetcpu_cond, &resetcpu_mtx); } pthread_mutex_unlock(&resetcpu_mtx); switch (how) { case VM_SUSPEND_RESET: exit(0); case VM_SUSPEND_POWEROFF: exit(1); case VM_SUSPEND_HALT: exit(2); case VM_SUSPEND_TRIPLEFAULT: exit(3); default: fprintf(stderr, "vmexit_suspend: invalid reason %d\n", how); exit(100); } return (0); /* NOTREACHED */ } static vmexit_handler_t handler[VM_EXITCODE_MAX] = { [VM_EXITCODE_INOUT] = vmexit_inout, [VM_EXITCODE_INOUT_STR] = vmexit_inout, [VM_EXITCODE_VMX] = vmexit_vmx, [VM_EXITCODE_SVM] = vmexit_svm, [VM_EXITCODE_BOGUS] = vmexit_bogus, [VM_EXITCODE_RDMSR] = vmexit_rdmsr, [VM_EXITCODE_WRMSR] = vmexit_wrmsr, [VM_EXITCODE_MTRAP] = vmexit_mtrap, [VM_EXITCODE_INST_EMUL] = vmexit_inst_emul, [VM_EXITCODE_SPINUP_AP] = vmexit_spinup_ap, [VM_EXITCODE_SUSPENDED] = vmexit_suspend, [VM_EXITCODE_TASK_SWITCH] = vmexit_task_switch, }; static void -vm_loop(struct vmctx *ctx, int vcpu, uint64_t rip) +vm_loop(struct vmctx *ctx, int vcpu, uint64_t startrip) { int error, rc, prevcpu; enum vm_exitcode exitcode; cpuset_t active_cpus; if (vcpumap[vcpu] != NULL) { error = pthread_setaffinity_np(pthread_self(), sizeof(cpuset_t), vcpumap[vcpu]); assert(error == 0); } error = vm_active_cpus(ctx, &active_cpus); assert(CPU_ISSET(vcpu, &active_cpus)); + error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, startrip); + assert(error == 0); + while (1) { - error = vm_run(ctx, vcpu, rip, &vmexit[vcpu]); + error = vm_run(ctx, vcpu, &vmexit[vcpu]); if (error != 0) break; prevcpu = vcpu; exitcode = vmexit[vcpu].exitcode; if (exitcode >= VM_EXITCODE_MAX || handler[exitcode] == NULL) { fprintf(stderr, "vm_loop: unexpected exitcode 0x%x\n", exitcode); exit(1); } rc = (*handler[exitcode])(ctx, &vmexit[vcpu], &vcpu); switch (rc) { case VMEXIT_CONTINUE: - rip = vmexit[vcpu].rip + vmexit[vcpu].inst_length; break; - case VMEXIT_RESTART: - rip = vmexit[vcpu].rip; - break; case VMEXIT_ABORT: abort(); default: exit(1); } } fprintf(stderr, "vm_run error %d, errno %d\n", error, errno); } static int num_vcpus_allowed(struct vmctx *ctx) { int tmp, error; error = vm_get_capability(ctx, BSP, VM_CAP_UNRESTRICTED_GUEST, &tmp); /* * The guest is allowed to spinup more than one processor only if the * UNRESTRICTED_GUEST capability is available. */ if (error == 0) return (VM_MAXCPU); else return (1); } void fbsdrun_set_capabilities(struct vmctx *ctx, int cpu) { int err, tmp; if (fbsdrun_vmexit_on_hlt()) { err = vm_get_capability(ctx, cpu, VM_CAP_HALT_EXIT, &tmp); if (err < 0) { fprintf(stderr, "VM exit on HLT not supported\n"); exit(1); } vm_set_capability(ctx, cpu, VM_CAP_HALT_EXIT, 1); if (cpu == BSP) handler[VM_EXITCODE_HLT] = vmexit_hlt; } if (fbsdrun_vmexit_on_pause()) { /* * pause exit support required for this mode */ err = vm_get_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, &tmp); if (err < 0) { fprintf(stderr, "SMP mux requested, no pause support\n"); exit(1); } vm_set_capability(ctx, cpu, VM_CAP_PAUSE_EXIT, 1); if (cpu == BSP) handler[VM_EXITCODE_PAUSE] = vmexit_pause; } if (x2apic_mode) err = vm_set_x2apic_state(ctx, cpu, X2APIC_ENABLED); else err = vm_set_x2apic_state(ctx, cpu, X2APIC_DISABLED); if (err) { fprintf(stderr, "Unable to set x2apic state (%d)\n", err); exit(1); } vm_set_capability(ctx, cpu, VM_CAP_ENABLE_INVPCID, 1); } int main(int argc, char *argv[]) { int c, error, gdb_port, err, bvmcons; int dump_guest_memory, max_vcpus, mptgen; + int rtc_localtime; struct vmctx *ctx; uint64_t rip; size_t memsize; bvmcons = 0; dump_guest_memory = 0; progname = basename(argv[0]); gdb_port = 0; guest_ncpus = 1; memsize = 256 * MB; mptgen = 1; + rtc_localtime = 1; - while ((c = getopt(argc, argv, "abehwxACHIPWYp:g:c:s:m:l:U:")) != -1) { + while ((c = getopt(argc, argv, "abehuwxACHIPWYp:g:c:s:m:l:U:")) != -1) { switch (c) { case 'a': x2apic_mode = 0; break; case 'A': acpi = 1; break; case 'b': bvmcons = 1; break; case 'p': if (pincpu_parse(optarg) != 0) { errx(EX_USAGE, "invalid vcpu pinning " "configuration '%s'", optarg); } break; case 'c': guest_ncpus = atoi(optarg); break; case 'C': dump_guest_memory = 1; break; case 'g': gdb_port = atoi(optarg); break; case 'l': if (lpc_device_parse(optarg) != 0) { errx(EX_USAGE, "invalid lpc device " "configuration '%s'", optarg); } break; case 's': if (pci_parse_slot(optarg) != 0) exit(1); else break; case 'm': error = vm_parse_memsize(optarg, &memsize); if (error) errx(EX_USAGE, "invalid memsize '%s'", optarg); break; case 'H': guest_vmexit_on_hlt = 1; break; case 'I': /* * The "-I" option was used to add an ioapic to the * virtual machine. * * An ioapic is now provided unconditionally for each * virtual machine and this option is now deprecated. */ break; case 'P': guest_vmexit_on_pause = 1; break; case 'e': strictio = 1; break; + case 'u': + rtc_localtime = 0; + break; case 'U': guest_uuid_str = optarg; break; case 'w': strictmsr = 0; break; case 'W': virtio_msix = 0; break; case 'x': x2apic_mode = 1; break; case 'Y': mptgen = 0; break; case 'h': usage(0); default: usage(1); } } argc -= optind; argv += optind; if (argc != 1) usage(1); vmname = argv[0]; ctx = vm_open(vmname); if (ctx == NULL) { perror("vm_open"); exit(1); } max_vcpus = num_vcpus_allowed(ctx); if (guest_ncpus > max_vcpus) { fprintf(stderr, "%d vCPUs requested but only %d available\n", guest_ncpus, max_vcpus); exit(1); } fbsdrun_set_capabilities(ctx, BSP); if (dump_guest_memory) vm_set_memflags(ctx, VM_MEM_F_INCORE); err = vm_setup_memory(ctx, memsize, VM_MMAP_ALL); if (err) { fprintf(stderr, "Unable to setup memory (%d)\n", err); exit(1); } error = init_msr(); if (error) { fprintf(stderr, "init_msr error %d", error); exit(1); } init_mem(); init_inout(); pci_irq_init(ctx); ioapic_init(ctx); - rtc_init(ctx); + rtc_init(ctx, rtc_localtime); sci_init(ctx); /* * Exit if a device emulation finds an error in it's initilization */ if (init_pci(ctx) != 0) exit(1); if (gdb_port != 0) init_dbgport(gdb_port); if (bvmcons) init_bvmcons(); error = vm_get_register(ctx, BSP, VM_REG_GUEST_RIP, &rip); assert(error == 0); /* * build the guest tables, MP etc. */ if (mptgen) { error = mptable_build(ctx, guest_ncpus); if (error) exit(1); } error = smbios_build(ctx); assert(error == 0); if (acpi) { error = acpi_build(ctx, guest_ncpus); assert(error == 0); } /* * Change the proc title to include the VM name. */ setproctitle("%s", vmname); /* * Add CPU 0 */ fbsdrun_addcpu(ctx, BSP, BSP, rip); /* * Head off to the main event dispatch loop */ mevent_dispatch(); exit(1); } Index: user/marcel/libvdsk/bhyve/bhyverun.h =================================================================== --- user/marcel/libvdsk/bhyve/bhyverun.h (revision 279574) +++ user/marcel/libvdsk/bhyve/bhyverun.h (revision 279575) @@ -1,56 +1,55 @@ /*- * Copyright (c) 2011 NetApp, 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 NETAPP, INC ``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 NETAPP, INC 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. * * $FreeBSD$ */ #ifndef _FBSDRUN_H_ #define _FBSDRUN_H_ #ifndef CTASSERT /* Allow lint to override */ #define CTASSERT(x) _CTASSERT(x, __LINE__) #define _CTASSERT(x, y) __CTASSERT(x, y) #define __CTASSERT(x, y) typedef char __assert ## y[(x) ? 1 : -1] #endif -#define VMEXIT_CONTINUE 1 /* continue from next instruction */ -#define VMEXIT_RESTART 2 /* restart current instruction */ -#define VMEXIT_ABORT 3 /* abort the vm run loop */ +#define VMEXIT_CONTINUE (0) +#define VMEXIT_ABORT (-1) struct vmctx; extern int guest_ncpus; extern char *guest_uuid_str; extern char *vmname; void *paddr_guest2host(struct vmctx *ctx, uintptr_t addr, size_t len); void fbsdrun_set_capabilities(struct vmctx *ctx, int cpu); void fbsdrun_addcpu(struct vmctx *ctx, int fromcpu, int newcpu, uint64_t rip); int fbsdrun_muxed(void); int fbsdrun_vmexit_on_hlt(void); int fbsdrun_vmexit_on_pause(void); int fbsdrun_disable_x2apic(void); int fbsdrun_virtio_msix(void); #endif Index: user/marcel/libvdsk/bhyve/inout.c =================================================================== --- user/marcel/libvdsk/bhyve/inout.c (revision 279574) +++ user/marcel/libvdsk/bhyve/inout.c (revision 279575) @@ -1,293 +1,297 @@ /*- * Copyright (c) 2011 NetApp, 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 NETAPP, INC ``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 NETAPP, INC 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. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include "bhyverun.h" #include "inout.h" SET_DECLARE(inout_port_set, struct inout_port); #define MAX_IOPORTS (1 << 16) #define VERIFY_IOPORT(port, size) \ assert((port) >= 0 && (size) > 0 && ((port) + (size)) <= MAX_IOPORTS) static struct { const char *name; int flags; inout_func_t handler; void *arg; } inout_handlers[MAX_IOPORTS]; static int default_inout(struct vmctx *ctx, int vcpu, int in, int port, int bytes, uint32_t *eax, void *arg) { if (in) { switch (bytes) { case 4: *eax = 0xffffffff; break; case 2: *eax = 0xffff; break; case 1: *eax = 0xff; break; } } return (0); } static void register_default_iohandler(int start, int size) { struct inout_port iop; VERIFY_IOPORT(start, size); bzero(&iop, sizeof(iop)); iop.name = "default"; iop.port = start; iop.size = size; iop.flags = IOPORT_F_INOUT | IOPORT_F_DEFAULT; iop.handler = default_inout; register_inout(&iop); } int emulate_inout(struct vmctx *ctx, int vcpu, struct vm_exit *vmexit, int strict) { int addrsize, bytes, flags, in, port, prot, rep; - uint32_t val; + uint32_t eax, val; inout_func_t handler; void *arg; int error, retval; enum vm_reg_name idxreg; uint64_t gla, index, iterations, count; struct vm_inout_str *vis; struct iovec iov[2]; bytes = vmexit->u.inout.bytes; in = vmexit->u.inout.in; port = vmexit->u.inout.port; assert(port < MAX_IOPORTS); assert(bytes == 1 || bytes == 2 || bytes == 4); handler = inout_handlers[port].handler; if (strict && handler == default_inout) return (-1); flags = inout_handlers[port].flags; arg = inout_handlers[port].arg; if (in) { if (!(flags & IOPORT_F_IN)) return (-1); } else { if (!(flags & IOPORT_F_OUT)) return (-1); } retval = 0; if (vmexit->u.inout.string) { vis = &vmexit->u.inout_str; rep = vis->inout.rep; addrsize = vis->addrsize; prot = in ? PROT_WRITE : PROT_READ; assert(addrsize == 2 || addrsize == 4 || addrsize == 8); /* Index register */ idxreg = in ? VM_REG_GUEST_RDI : VM_REG_GUEST_RSI; index = vis->index & vie_size2mask(addrsize); /* Count register */ count = vis->count & vie_size2mask(addrsize); /* Limit number of back-to-back in/out emulations to 16 */ iterations = MIN(count, 16); while (iterations > 0) { assert(retval == 0); if (vie_calculate_gla(vis->paging.cpu_mode, vis->seg_name, &vis->seg_desc, index, bytes, addrsize, prot, &gla)) { vm_inject_gp(ctx, vcpu); break; } error = vm_copy_setup(ctx, vcpu, &vis->paging, gla, bytes, prot, iov, nitems(iov)); if (error == -1) { retval = -1; /* Unrecoverable error */ break; } else if (error == 1) { retval = 0; /* Resume guest to handle fault */ break; } if (vie_alignment_check(vis->paging.cpl, bytes, vis->cr0, vis->rflags, gla)) { vm_inject_ac(ctx, vcpu, 0); break; } val = 0; if (!in) vm_copyin(ctx, vcpu, iov, &val, bytes); retval = handler(ctx, vcpu, in, port, bytes, &val, arg); if (retval != 0) break; if (in) vm_copyout(ctx, vcpu, &val, iov, bytes); /* Update index */ if (vis->rflags & PSL_D) index -= bytes; else index += bytes; count--; iterations--; } /* Update index register */ error = vie_update_register(ctx, vcpu, idxreg, index, addrsize); assert(error == 0); /* * Update count register only if the instruction had a repeat * prefix. */ if (rep) { error = vie_update_register(ctx, vcpu, VM_REG_GUEST_RCX, count, addrsize); assert(error == 0); } /* Restart the instruction if more iterations remain */ - if (retval == 0 && count != 0) - vmexit->inst_length = 0; - } else { - if (!in) { - val = vmexit->u.inout.eax & vie_size2mask(bytes); + if (retval == 0 && count != 0) { + error = vm_restart_instruction(ctx, vcpu); + assert(error == 0); } + } else { + eax = vmexit->u.inout.eax; + val = eax & vie_size2mask(bytes); retval = handler(ctx, vcpu, in, port, bytes, &val, arg); if (retval == 0 && in) { - vmexit->u.inout.eax &= ~vie_size2mask(bytes); - vmexit->u.inout.eax |= val & vie_size2mask(bytes); + eax &= ~vie_size2mask(bytes); + eax |= val & vie_size2mask(bytes); + error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RAX, + eax); + assert(error == 0); } } return (retval); } void init_inout(void) { struct inout_port **iopp, *iop; /* * Set up the default handler for all ports */ register_default_iohandler(0, MAX_IOPORTS); /* * Overwrite with specified handlers */ SET_FOREACH(iopp, inout_port_set) { iop = *iopp; assert(iop->port < MAX_IOPORTS); inout_handlers[iop->port].name = iop->name; inout_handlers[iop->port].flags = iop->flags; inout_handlers[iop->port].handler = iop->handler; inout_handlers[iop->port].arg = NULL; } } int register_inout(struct inout_port *iop) { int i; VERIFY_IOPORT(iop->port, iop->size); /* * Verify that the new registration is not overwriting an already * allocated i/o range. */ if ((iop->flags & IOPORT_F_DEFAULT) == 0) { for (i = iop->port; i < iop->port + iop->size; i++) { if ((inout_handlers[i].flags & IOPORT_F_DEFAULT) == 0) return (-1); } } for (i = iop->port; i < iop->port + iop->size; i++) { inout_handlers[i].name = iop->name; inout_handlers[i].flags = iop->flags; inout_handlers[i].handler = iop->handler; inout_handlers[i].arg = iop->arg; } return (0); } int unregister_inout(struct inout_port *iop) { VERIFY_IOPORT(iop->port, iop->size); assert(inout_handlers[iop->port].name == iop->name); register_default_iohandler(iop->port, iop->size); return (0); } Index: user/marcel/libvdsk/bhyve/pci_ahci.c =================================================================== --- user/marcel/libvdsk/bhyve/pci_ahci.c (revision 279574) +++ user/marcel/libvdsk/bhyve/pci_ahci.c (revision 279575) @@ -1,2009 +1,2010 @@ /*- * Copyright (c) 2013 Zhixiang Yu * 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 AUTHOR ``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. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "bhyverun.h" #include "pci_emul.h" #include "ahci.h" #include "block_if.h" #define MAX_PORTS 6 /* Intel ICH8 AHCI supports 6 ports */ #define PxSIG_ATA 0x00000101 /* ATA drive */ #define PxSIG_ATAPI 0xeb140101 /* ATAPI drive */ enum sata_fis_type { FIS_TYPE_REGH2D = 0x27, /* Register FIS - host to device */ FIS_TYPE_REGD2H = 0x34, /* Register FIS - device to host */ FIS_TYPE_DMAACT = 0x39, /* DMA activate FIS - device to host */ FIS_TYPE_DMASETUP = 0x41, /* DMA setup FIS - bidirectional */ FIS_TYPE_DATA = 0x46, /* Data FIS - bidirectional */ FIS_TYPE_BIST = 0x58, /* BIST activate FIS - bidirectional */ FIS_TYPE_PIOSETUP = 0x5F, /* PIO setup FIS - device to host */ FIS_TYPE_SETDEVBITS = 0xA1, /* Set dev bits FIS - device to host */ }; /* * SCSI opcodes */ #define TEST_UNIT_READY 0x00 #define REQUEST_SENSE 0x03 #define INQUIRY 0x12 #define START_STOP_UNIT 0x1B #define PREVENT_ALLOW 0x1E #define READ_CAPACITY 0x25 #define READ_10 0x28 #define POSITION_TO_ELEMENT 0x2B #define READ_TOC 0x43 #define GET_EVENT_STATUS_NOTIFICATION 0x4A #define MODE_SENSE_10 0x5A #define READ_12 0xA8 #define READ_CD 0xBE /* * SCSI mode page codes */ #define MODEPAGE_RW_ERROR_RECOVERY 0x01 #define MODEPAGE_CD_CAPABILITIES 0x2A /* * ATA commands */ #define ATA_SF_ENAB_SATA_SF 0x10 #define ATA_SATA_SF_AN 0x05 #define ATA_SF_DIS_SATA_SF 0x90 /* * Debug printf */ #ifdef AHCI_DEBUG static FILE *dbg; #define DPRINTF(format, arg...) do{fprintf(dbg, format, ##arg);fflush(dbg);}while(0) #else #define DPRINTF(format, arg...) #endif #define WPRINTF(format, arg...) printf(format, ##arg) struct ahci_ioreq { struct blockif_req io_req; struct ahci_port *io_pr; STAILQ_ENTRY(ahci_ioreq) io_flist; TAILQ_ENTRY(ahci_ioreq) io_blist; uint8_t *cfis; uint32_t len; uint32_t done; int slot; int prdtl; }; struct ahci_port { struct blockif_ctxt *bctx; struct pci_ahci_softc *pr_sc; uint8_t *cmd_lst; uint8_t *rfis; int atapi; int reset; int mult_sectors; uint8_t xfermode; uint8_t sense_key; uint8_t asc; uint32_t pending; uint32_t clb; uint32_t clbu; uint32_t fb; uint32_t fbu; uint32_t is; uint32_t ie; uint32_t cmd; uint32_t unused0; uint32_t tfd; uint32_t sig; uint32_t ssts; uint32_t sctl; uint32_t serr; uint32_t sact; uint32_t ci; uint32_t sntf; uint32_t fbs; /* * i/o request info */ struct ahci_ioreq *ioreq; int ioqsz; STAILQ_HEAD(ahci_fhead, ahci_ioreq) iofhd; TAILQ_HEAD(ahci_bhead, ahci_ioreq) iobhd; }; struct ahci_cmd_hdr { uint16_t flags; uint16_t prdtl; uint32_t prdbc; uint64_t ctba; uint32_t reserved[4]; }; struct ahci_prdt_entry { uint64_t dba; uint32_t reserved; #define DBCMASK 0x3fffff uint32_t dbc; }; struct pci_ahci_softc { struct pci_devinst *asc_pi; pthread_mutex_t mtx; int ports; uint32_t cap; uint32_t ghc; uint32_t is; uint32_t pi; uint32_t vs; uint32_t ccc_ctl; uint32_t ccc_pts; uint32_t em_loc; uint32_t em_ctl; uint32_t cap2; uint32_t bohc; uint32_t lintr; struct ahci_port port[MAX_PORTS]; }; #define ahci_ctx(sc) ((sc)->asc_pi->pi_vmctx) static inline void lba_to_msf(uint8_t *buf, int lba) { lba += 150; buf[0] = (lba / 75) / 60; buf[1] = (lba / 75) % 60; buf[2] = lba % 75; } /* * generate HBA intr depending on whether or not ports within * the controller have an interrupt pending. */ static void ahci_generate_intr(struct pci_ahci_softc *sc) { struct pci_devinst *pi; int i; pi = sc->asc_pi; for (i = 0; i < sc->ports; i++) { struct ahci_port *pr; pr = &sc->port[i]; if (pr->is & pr->ie) sc->is |= (1 << i); } DPRINTF("%s %x\n", __func__, sc->is); if (sc->is && (sc->ghc & AHCI_GHC_IE)) { if (pci_msi_enabled(pi)) { /* * Generate an MSI interrupt on every edge */ pci_generate_msi(pi, 0); } else if (!sc->lintr) { /* * Only generate a pin-based interrupt if one wasn't * in progress */ sc->lintr = 1; pci_lintr_assert(pi); } } else if (sc->lintr) { /* * No interrupts: deassert pin-based signal if it had * been asserted */ pci_lintr_deassert(pi); sc->lintr = 0; } } static void ahci_write_fis(struct ahci_port *p, enum sata_fis_type ft, uint8_t *fis) { int offset, len, irq; if (p->rfis == NULL || !(p->cmd & AHCI_P_CMD_FRE)) return; switch (ft) { case FIS_TYPE_REGD2H: offset = 0x40; len = 20; irq = AHCI_P_IX_DHR; break; case FIS_TYPE_SETDEVBITS: offset = 0x58; len = 8; irq = AHCI_P_IX_SDB; break; case FIS_TYPE_PIOSETUP: offset = 0x20; len = 20; irq = 0; break; default: WPRINTF("unsupported fis type %d\n", ft); return; } memcpy(p->rfis + offset, fis, len); if (irq) { p->is |= irq; ahci_generate_intr(p->pr_sc); } } static void ahci_write_fis_piosetup(struct ahci_port *p) { uint8_t fis[20]; memset(fis, 0, sizeof(fis)); fis[0] = FIS_TYPE_PIOSETUP; ahci_write_fis(p, FIS_TYPE_PIOSETUP, fis); } static void ahci_write_fis_sdb(struct ahci_port *p, int slot, uint32_t tfd) { uint8_t fis[8]; uint8_t error; error = (tfd >> 8) & 0xff; memset(fis, 0, sizeof(fis)); fis[0] = error; fis[2] = tfd & 0x77; *(uint32_t *)(fis + 4) = (1 << slot); if (fis[2] & ATA_S_ERROR) p->is |= AHCI_P_IX_TFE; p->tfd = tfd; ahci_write_fis(p, FIS_TYPE_SETDEVBITS, fis); } static void ahci_write_fis_d2h(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t tfd) { uint8_t fis[20]; uint8_t error; error = (tfd >> 8) & 0xff; memset(fis, 0, sizeof(fis)); fis[0] = FIS_TYPE_REGD2H; fis[1] = (1 << 6); fis[2] = tfd & 0xff; fis[3] = error; fis[4] = cfis[4]; fis[5] = cfis[5]; fis[6] = cfis[6]; fis[7] = cfis[7]; fis[8] = cfis[8]; fis[9] = cfis[9]; fis[10] = cfis[10]; fis[11] = cfis[11]; fis[12] = cfis[12]; fis[13] = cfis[13]; if (fis[2] & ATA_S_ERROR) p->is |= AHCI_P_IX_TFE; else p->ci &= ~(1 << slot); p->tfd = tfd; ahci_write_fis(p, FIS_TYPE_REGD2H, fis); } static void ahci_write_reset_fis_d2h(struct ahci_port *p) { uint8_t fis[20]; memset(fis, 0, sizeof(fis)); fis[0] = FIS_TYPE_REGD2H; fis[3] = 1; fis[4] = 1; if (p->atapi) { fis[5] = 0x14; fis[6] = 0xeb; } fis[12] = 1; ahci_write_fis(p, FIS_TYPE_REGD2H, fis); } static void ahci_check_stopped(struct ahci_port *p) { /* * If we are no longer processing the command list and nothing * is in-flight, clear the running bit, the current command * slot, the command issue and active bits. */ if (!(p->cmd & AHCI_P_CMD_ST)) { if (p->pending == 0) { p->cmd &= ~(AHCI_P_CMD_CR | AHCI_P_CMD_CCS_MASK); p->ci = 0; p->sact = 0; } } } static void ahci_port_stop(struct ahci_port *p) { struct ahci_ioreq *aior; uint8_t *cfis; int slot; int ncq; int error; assert(pthread_mutex_isowned_np(&p->pr_sc->mtx)); TAILQ_FOREACH(aior, &p->iobhd, io_blist) { /* * Try to cancel the outstanding blockif request. */ error = blockif_cancel(p->bctx, &aior->io_req); if (error != 0) continue; slot = aior->slot; cfis = aior->cfis; if (cfis[2] == ATA_WRITE_FPDMA_QUEUED || cfis[2] == ATA_READ_FPDMA_QUEUED) ncq = 1; if (ncq) p->sact &= ~(1 << slot); else p->ci &= ~(1 << slot); /* * This command is now done. */ p->pending &= ~(1 << slot); /* * Delete the blockif request from the busy list */ TAILQ_REMOVE(&p->iobhd, aior, io_blist); /* * Move the blockif request back to the free list */ STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist); } ahci_check_stopped(p); } static void ahci_port_reset(struct ahci_port *pr) { pr->sctl = 0; pr->serr = 0; pr->sact = 0; pr->xfermode = ATA_UDMA6; pr->mult_sectors = 128; if (!pr->bctx) { pr->ssts = ATA_SS_DET_NO_DEVICE; pr->sig = 0xFFFFFFFF; pr->tfd = 0x7F; return; } pr->ssts = ATA_SS_DET_PHY_ONLINE | ATA_SS_SPD_GEN2 | ATA_SS_IPM_ACTIVE; pr->tfd = (1 << 8) | ATA_S_DSC | ATA_S_DMA; if (!pr->atapi) { pr->sig = PxSIG_ATA; pr->tfd |= ATA_S_READY; } else pr->sig = PxSIG_ATAPI; ahci_write_reset_fis_d2h(pr); } static void ahci_reset(struct pci_ahci_softc *sc) { int i; sc->ghc = AHCI_GHC_AE; sc->is = 0; if (sc->lintr) { pci_lintr_deassert(sc->asc_pi); sc->lintr = 0; } for (i = 0; i < sc->ports; i++) { sc->port[i].ie = 0; sc->port[i].is = 0; ahci_port_reset(&sc->port[i]); } } static void ata_string(uint8_t *dest, const char *src, int len) { int i; for (i = 0; i < len; i++) { if (*src) dest[i ^ 1] = *src++; else dest[i ^ 1] = ' '; } } static void atapi_string(uint8_t *dest, const char *src, int len) { int i; for (i = 0; i < len; i++) { if (*src) dest[i] = *src++; else dest[i] = ' '; } } static void ahci_handle_dma(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t done, int seek) { struct ahci_ioreq *aior; struct blockif_req *breq; struct pci_ahci_softc *sc; struct ahci_prdt_entry *prdt; struct ahci_cmd_hdr *hdr; uint64_t lba; uint32_t len; int i, err, iovcnt, ncq, readop; sc = p->pr_sc; prdt = (struct ahci_prdt_entry *)(cfis + 0x80); hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE); ncq = 0; readop = 1; prdt += seek; if (cfis[2] == ATA_WRITE_DMA || cfis[2] == ATA_WRITE_DMA48 || cfis[2] == ATA_WRITE_FPDMA_QUEUED) readop = 0; if (cfis[2] == ATA_WRITE_FPDMA_QUEUED || cfis[2] == ATA_READ_FPDMA_QUEUED) { lba = ((uint64_t)cfis[10] << 40) | ((uint64_t)cfis[9] << 32) | ((uint64_t)cfis[8] << 24) | ((uint64_t)cfis[6] << 16) | ((uint64_t)cfis[5] << 8) | cfis[4]; len = cfis[11] << 8 | cfis[3]; if (!len) len = 65536; ncq = 1; } else if (cfis[2] == ATA_READ_DMA48 || cfis[2] == ATA_WRITE_DMA48) { lba = ((uint64_t)cfis[10] << 40) | ((uint64_t)cfis[9] << 32) | ((uint64_t)cfis[8] << 24) | ((uint64_t)cfis[6] << 16) | ((uint64_t)cfis[5] << 8) | cfis[4]; len = cfis[13] << 8 | cfis[12]; if (!len) len = 65536; } else { lba = ((cfis[7] & 0xf) << 24) | (cfis[6] << 16) | (cfis[5] << 8) | cfis[4]; len = cfis[12]; if (!len) len = 256; } lba *= blockif_sectsz(p->bctx); len *= blockif_sectsz(p->bctx); /* * Pull request off free list */ aior = STAILQ_FIRST(&p->iofhd); assert(aior != NULL); STAILQ_REMOVE_HEAD(&p->iofhd, io_flist); aior->cfis = cfis; aior->slot = slot; aior->len = len; aior->done = done; breq = &aior->io_req; breq->br_offset = lba + done; iovcnt = hdr->prdtl - seek; if (iovcnt > BLOCKIF_IOV_MAX) { aior->prdtl = iovcnt - BLOCKIF_IOV_MAX; iovcnt = BLOCKIF_IOV_MAX; } else aior->prdtl = 0; breq->br_iovcnt = iovcnt; /* * Mark this command in-flight. */ p->pending |= 1 << slot; /* * Stuff request onto busy list */ TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist); /* * Build up the iovec based on the prdt */ for (i = 0; i < iovcnt; i++) { uint32_t dbcsz; dbcsz = (prdt->dbc & DBCMASK) + 1; breq->br_iov[i].iov_base = paddr_guest2host(ahci_ctx(sc), prdt->dba, dbcsz); breq->br_iov[i].iov_len = dbcsz; aior->done += dbcsz; prdt++; } if (readop) err = blockif_read(p->bctx, breq); else err = blockif_write(p->bctx, breq); assert(err == 0); if (ncq) p->ci &= ~(1 << slot); } static void ahci_handle_flush(struct ahci_port *p, int slot, uint8_t *cfis) { struct ahci_ioreq *aior; struct blockif_req *breq; int err; /* * Pull request off free list */ aior = STAILQ_FIRST(&p->iofhd); assert(aior != NULL); STAILQ_REMOVE_HEAD(&p->iofhd, io_flist); aior->cfis = cfis; aior->slot = slot; aior->len = 0; aior->done = 0; aior->prdtl = 0; breq = &aior->io_req; /* * Mark this command in-flight. */ p->pending |= 1 << slot; /* * Stuff request onto busy list */ TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist); err = blockif_flush(p->bctx, breq); assert(err == 0); } static inline void write_prdt(struct ahci_port *p, int slot, uint8_t *cfis, void *buf, int size) { struct ahci_cmd_hdr *hdr; struct ahci_prdt_entry *prdt; void *from; int i, len; hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE); len = size; from = buf; prdt = (struct ahci_prdt_entry *)(cfis + 0x80); for (i = 0; i < hdr->prdtl && len; i++) { uint8_t *ptr; uint32_t dbcsz; int sublen; dbcsz = (prdt->dbc & DBCMASK) + 1; ptr = paddr_guest2host(ahci_ctx(p->pr_sc), prdt->dba, dbcsz); sublen = len < dbcsz ? len : dbcsz; memcpy(ptr, from, sublen); len -= sublen; from += sublen; prdt++; } hdr->prdbc = size - len; } static void handle_identify(struct ahci_port *p, int slot, uint8_t *cfis) { struct ahci_cmd_hdr *hdr; hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE); if (p->atapi || hdr->prdtl == 0) { p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR; p->is |= AHCI_P_IX_TFE; } else { uint16_t buf[256]; uint64_t sectors; uint16_t cyl; uint8_t sech, heads; sectors = blockif_size(p->bctx) / blockif_sectsz(p->bctx); blockif_chs(p->bctx, &cyl, &heads, &sech); memset(buf, 0, sizeof(buf)); buf[0] = 0x0040; buf[1] = cyl; buf[3] = heads; buf[6] = sech; /* TODO emulate different serial? */ ata_string((uint8_t *)(buf+10), "123456", 20); ata_string((uint8_t *)(buf+23), "001", 8); ata_string((uint8_t *)(buf+27), "BHYVE SATA DISK", 40); buf[47] = (0x8000 | 128); buf[48] = 0x1; buf[49] = (1 << 8 | 1 << 9 | 1 << 11); buf[50] = (1 << 14); buf[53] = (1 << 1 | 1 << 2); if (p->mult_sectors) buf[59] = (0x100 | p->mult_sectors); buf[60] = sectors; buf[61] = (sectors >> 16); buf[63] = 0x7; if (p->xfermode & ATA_WDMA0) buf[63] |= (1 << ((p->xfermode & 7) + 8)); buf[64] = 0x3; buf[65] = 100; buf[66] = 100; buf[67] = 100; buf[68] = 100; buf[75] = 31; buf[76] = (1 << 8 | 1 << 2); buf[80] = 0x1f0; buf[81] = 0x28; buf[82] = (1 << 5 | 1 << 14); buf[83] = (1 << 10 | 1 << 12 | 1 << 13 | 1 << 14); buf[84] = (1 << 14); buf[85] = (1 << 5 | 1 << 14); buf[86] = (1 << 10 | 1 << 12 | 1 << 13); buf[87] = (1 << 14); buf[88] = 0x7f; if (p->xfermode & ATA_UDMA0) buf[88] |= (1 << ((p->xfermode & 7) + 8)); buf[93] = (1 | 1 <<14); buf[100] = sectors; buf[101] = (sectors >> 16); buf[102] = (sectors >> 32); buf[103] = (sectors >> 48); ahci_write_fis_piosetup(p); write_prdt(p, slot, cfis, (void *)buf, sizeof(buf)); p->tfd = ATA_S_DSC | ATA_S_READY; p->is |= AHCI_P_IX_DP; p->ci &= ~(1 << slot); } ahci_generate_intr(p->pr_sc); } static void handle_atapi_identify(struct ahci_port *p, int slot, uint8_t *cfis) { if (!p->atapi) { p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR; p->is |= AHCI_P_IX_TFE; } else { uint16_t buf[256]; memset(buf, 0, sizeof(buf)); buf[0] = (2 << 14 | 5 << 8 | 1 << 7 | 2 << 5); /* TODO emulate different serial? */ ata_string((uint8_t *)(buf+10), "123456", 20); ata_string((uint8_t *)(buf+23), "001", 8); ata_string((uint8_t *)(buf+27), "BHYVE SATA DVD ROM", 40); buf[49] = (1 << 9 | 1 << 8); buf[50] = (1 << 14 | 1); buf[53] = (1 << 2 | 1 << 1); buf[62] = 0x3f; buf[63] = 7; buf[64] = 3; buf[65] = 100; buf[66] = 100; buf[67] = 100; buf[68] = 100; buf[76] = (1 << 2 | 1 << 1); buf[78] = (1 << 5); buf[80] = (0x1f << 4); buf[82] = (1 << 4); buf[83] = (1 << 14); buf[84] = (1 << 14); buf[85] = (1 << 4); buf[87] = (1 << 14); buf[88] = (1 << 14 | 0x7f); ahci_write_fis_piosetup(p); write_prdt(p, slot, cfis, (void *)buf, sizeof(buf)); p->tfd = ATA_S_DSC | ATA_S_READY; p->is |= AHCI_P_IX_DHR; p->ci &= ~(1 << slot); } ahci_generate_intr(p->pr_sc); } static void atapi_inquiry(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t buf[36]; uint8_t *acmd; int len; acmd = cfis + 0x40; buf[0] = 0x05; buf[1] = 0x80; buf[2] = 0x00; buf[3] = 0x21; buf[4] = 31; buf[5] = 0; buf[6] = 0; buf[7] = 0; atapi_string(buf + 8, "BHYVE", 8); atapi_string(buf + 16, "BHYVE DVD-ROM", 16); atapi_string(buf + 32, "001", 4); len = sizeof(buf); if (len > acmd[4]) len = acmd[4]; cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; write_prdt(p, slot, cfis, buf, len); ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); } static void atapi_read_capacity(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t buf[8]; uint64_t sectors; sectors = blockif_size(p->bctx) / 2048; be32enc(buf, sectors - 1); be32enc(buf + 4, 2048); cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; write_prdt(p, slot, cfis, buf, sizeof(buf)); ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); } static void atapi_read_toc(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t *acmd; uint8_t format; int len; acmd = cfis + 0x40; len = be16dec(acmd + 7); format = acmd[9] >> 6; switch (format) { case 0: { int msf, size; uint64_t sectors; uint8_t start_track, buf[20], *bp; msf = (acmd[1] >> 1) & 1; start_track = acmd[6]; if (start_track > 1 && start_track != 0xaa) { uint32_t tfd; p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x24; tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR; cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, tfd); return; } bp = buf + 2; *bp++ = 1; *bp++ = 1; if (start_track <= 1) { *bp++ = 0; *bp++ = 0x14; *bp++ = 1; *bp++ = 0; if (msf) { *bp++ = 0; lba_to_msf(bp, 0); bp += 3; } else { *bp++ = 0; *bp++ = 0; *bp++ = 0; *bp++ = 0; } } *bp++ = 0; *bp++ = 0x14; *bp++ = 0xaa; *bp++ = 0; sectors = blockif_size(p->bctx) / blockif_sectsz(p->bctx); sectors >>= 2; if (msf) { *bp++ = 0; lba_to_msf(bp, sectors); bp += 3; } else { be32enc(bp, sectors); bp += 4; } size = bp - buf; be16enc(buf, size - 2); if (len > size) len = size; write_prdt(p, slot, cfis, buf, len); cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); break; } case 1: { uint8_t buf[12]; memset(buf, 0, sizeof(buf)); buf[1] = 0xa; buf[2] = 0x1; buf[3] = 0x1; if (len > sizeof(buf)) len = sizeof(buf); write_prdt(p, slot, cfis, buf, len); cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); break; } case 2: { int msf, size; uint64_t sectors; uint8_t start_track, *bp, buf[50]; msf = (acmd[1] >> 1) & 1; start_track = acmd[6]; bp = buf + 2; *bp++ = 1; *bp++ = 1; *bp++ = 1; *bp++ = 0x14; *bp++ = 0; *bp++ = 0xa0; *bp++ = 0; *bp++ = 0; *bp++ = 0; *bp++ = 0; *bp++ = 1; *bp++ = 0; *bp++ = 0; *bp++ = 1; *bp++ = 0x14; *bp++ = 0; *bp++ = 0xa1; *bp++ = 0; *bp++ = 0; *bp++ = 0; *bp++ = 0; *bp++ = 1; *bp++ = 0; *bp++ = 0; *bp++ = 1; *bp++ = 0x14; *bp++ = 0; *bp++ = 0xa2; *bp++ = 0; *bp++ = 0; *bp++ = 0; sectors = blockif_size(p->bctx) / blockif_sectsz(p->bctx); sectors >>= 2; if (msf) { *bp++ = 0; lba_to_msf(bp, sectors); bp += 3; } else { be32enc(bp, sectors); bp += 4; } *bp++ = 1; *bp++ = 0x14; *bp++ = 0; *bp++ = 1; *bp++ = 0; *bp++ = 0; *bp++ = 0; if (msf) { *bp++ = 0; lba_to_msf(bp, 0); bp += 3; } else { *bp++ = 0; *bp++ = 0; *bp++ = 0; *bp++ = 0; } size = bp - buf; be16enc(buf, size - 2); if (len > size) len = size; write_prdt(p, slot, cfis, buf, len); cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); break; } default: { uint32_t tfd; p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x24; tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR; cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, tfd); break; } } } static void atapi_read(struct ahci_port *p, int slot, uint8_t *cfis, uint32_t done, int seek) { struct ahci_ioreq *aior; struct ahci_cmd_hdr *hdr; struct ahci_prdt_entry *prdt; struct blockif_req *breq; struct pci_ahci_softc *sc; uint8_t *acmd; uint64_t lba; uint32_t len; int i, err, iovcnt; sc = p->pr_sc; acmd = cfis + 0x40; hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE); prdt = (struct ahci_prdt_entry *)(cfis + 0x80); prdt += seek; lba = be32dec(acmd + 2); if (acmd[0] == READ_10) len = be16dec(acmd + 7); else len = be32dec(acmd + 6); if (len == 0) { cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); } lba *= 2048; len *= 2048; /* * Pull request off free list */ aior = STAILQ_FIRST(&p->iofhd); assert(aior != NULL); STAILQ_REMOVE_HEAD(&p->iofhd, io_flist); aior->cfis = cfis; aior->slot = slot; aior->len = len; aior->done = done; breq = &aior->io_req; breq->br_offset = lba + done; iovcnt = hdr->prdtl - seek; if (iovcnt > BLOCKIF_IOV_MAX) { aior->prdtl = iovcnt - BLOCKIF_IOV_MAX; iovcnt = BLOCKIF_IOV_MAX; } else aior->prdtl = 0; breq->br_iovcnt = iovcnt; /* * Mark this command in-flight. */ p->pending |= 1 << slot; /* * Stuff request onto busy list */ TAILQ_INSERT_HEAD(&p->iobhd, aior, io_blist); /* * Build up the iovec based on the prdt */ for (i = 0; i < iovcnt; i++) { uint32_t dbcsz; dbcsz = (prdt->dbc & DBCMASK) + 1; breq->br_iov[i].iov_base = paddr_guest2host(ahci_ctx(sc), prdt->dba, dbcsz); breq->br_iov[i].iov_len = dbcsz; aior->done += dbcsz; prdt++; } err = blockif_read(p->bctx, breq); assert(err == 0); } static void atapi_request_sense(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t buf[64]; uint8_t *acmd; int len; acmd = cfis + 0x40; len = acmd[4]; if (len > sizeof(buf)) len = sizeof(buf); memset(buf, 0, len); buf[0] = 0x70 | (1 << 7); buf[2] = p->sense_key; buf[7] = 10; buf[12] = p->asc; write_prdt(p, slot, cfis, buf, len); cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); } static void atapi_start_stop_unit(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t *acmd = cfis + 0x40; uint32_t tfd; switch (acmd[4] & 3) { case 0: case 1: case 3: cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; tfd = ATA_S_READY | ATA_S_DSC; break; case 2: /* TODO eject media */ cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x53; tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR; break; } ahci_write_fis_d2h(p, slot, cfis, tfd); } static void atapi_mode_sense(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t *acmd; uint32_t tfd; uint8_t pc, code; int len; acmd = cfis + 0x40; len = be16dec(acmd + 7); pc = acmd[2] >> 6; code = acmd[2] & 0x3f; switch (pc) { case 0: switch (code) { case MODEPAGE_RW_ERROR_RECOVERY: { uint8_t buf[16]; if (len > sizeof(buf)) len = sizeof(buf); memset(buf, 0, sizeof(buf)); be16enc(buf, 16 - 2); buf[2] = 0x70; buf[8] = 0x01; buf[9] = 16 - 10; buf[11] = 0x05; write_prdt(p, slot, cfis, buf, len); tfd = ATA_S_READY | ATA_S_DSC; break; } case MODEPAGE_CD_CAPABILITIES: { uint8_t buf[30]; if (len > sizeof(buf)) len = sizeof(buf); memset(buf, 0, sizeof(buf)); be16enc(buf, 30 - 2); buf[2] = 0x70; buf[8] = 0x2A; buf[9] = 30 - 10; buf[10] = 0x08; buf[12] = 0x71; be16enc(&buf[18], 2); be16enc(&buf[20], 512); write_prdt(p, slot, cfis, buf, len); tfd = ATA_S_READY | ATA_S_DSC; break; } default: goto error; break; } break; case 3: p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x39; tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR; break; error: case 1: case 2: p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x24; tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR; break; } cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, tfd); } static void atapi_get_event_status_notification(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t *acmd; uint32_t tfd; acmd = cfis + 0x40; /* we don't support asynchronous operation */ if (!(acmd[1] & 1)) { p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x24; tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR; } else { uint8_t buf[8]; int len; len = be16dec(acmd + 7); if (len > sizeof(buf)) len = sizeof(buf); memset(buf, 0, sizeof(buf)); be16enc(buf, 8 - 2); buf[2] = 0x04; buf[3] = 0x10; buf[5] = 0x02; write_prdt(p, slot, cfis, buf, len); tfd = ATA_S_READY | ATA_S_DSC; } cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, tfd); } static void handle_packet_cmd(struct ahci_port *p, int slot, uint8_t *cfis) { uint8_t *acmd; acmd = cfis + 0x40; #ifdef AHCI_DEBUG { int i; DPRINTF("ACMD:"); for (i = 0; i < 16; i++) DPRINTF("%02x ", acmd[i]); DPRINTF("\n"); } #endif switch (acmd[0]) { case TEST_UNIT_READY: cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); break; case INQUIRY: atapi_inquiry(p, slot, cfis); break; case READ_CAPACITY: atapi_read_capacity(p, slot, cfis); break; case PREVENT_ALLOW: /* TODO */ cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); break; case READ_TOC: atapi_read_toc(p, slot, cfis); break; case READ_10: case READ_12: atapi_read(p, slot, cfis, 0, 0); break; case REQUEST_SENSE: atapi_request_sense(p, slot, cfis); break; case START_STOP_UNIT: atapi_start_stop_unit(p, slot, cfis); break; case MODE_SENSE_10: atapi_mode_sense(p, slot, cfis); break; case GET_EVENT_STATUS_NOTIFICATION: atapi_get_event_status_notification(p, slot, cfis); break; default: cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x20; ahci_write_fis_d2h(p, slot, cfis, (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR); break; } } static void ahci_handle_cmd(struct ahci_port *p, int slot, uint8_t *cfis) { switch (cfis[2]) { case ATA_ATA_IDENTIFY: handle_identify(p, slot, cfis); break; case ATA_SETFEATURES: { switch (cfis[3]) { case ATA_SF_ENAB_SATA_SF: switch (cfis[12]) { case ATA_SATA_SF_AN: p->tfd = ATA_S_DSC | ATA_S_READY; break; default: p->tfd = ATA_S_ERROR | ATA_S_READY; p->tfd |= (ATA_ERROR_ABORT << 8); break; } break; case ATA_SF_ENAB_WCACHE: case ATA_SF_DIS_WCACHE: case ATA_SF_ENAB_RCACHE: case ATA_SF_DIS_RCACHE: p->tfd = ATA_S_DSC | ATA_S_READY; break; case ATA_SF_SETXFER: { switch (cfis[12] & 0xf8) { case ATA_PIO: case ATA_PIO0: break; case ATA_WDMA0: case ATA_UDMA0: p->xfermode = (cfis[12] & 0x7); break; } p->tfd = ATA_S_DSC | ATA_S_READY; break; } default: p->tfd = ATA_S_ERROR | ATA_S_READY; p->tfd |= (ATA_ERROR_ABORT << 8); break; } ahci_write_fis_d2h(p, slot, cfis, p->tfd); break; } case ATA_SET_MULTI: if (cfis[12] != 0 && (cfis[12] > 128 || (cfis[12] & (cfis[12] - 1)))) { p->tfd = ATA_S_ERROR | ATA_S_READY; p->tfd |= (ATA_ERROR_ABORT << 8); } else { p->mult_sectors = cfis[12]; p->tfd = ATA_S_DSC | ATA_S_READY; } p->is |= AHCI_P_IX_DP; p->ci &= ~(1 << slot); ahci_generate_intr(p->pr_sc); break; case ATA_READ_DMA: case ATA_WRITE_DMA: case ATA_READ_DMA48: case ATA_WRITE_DMA48: case ATA_READ_FPDMA_QUEUED: case ATA_WRITE_FPDMA_QUEUED: ahci_handle_dma(p, slot, cfis, 0, 0); break; case ATA_FLUSHCACHE: case ATA_FLUSHCACHE48: ahci_handle_flush(p, slot, cfis); break; case ATA_STANDBY_CMD: break; case ATA_NOP: case ATA_STANDBY_IMMEDIATE: case ATA_IDLE_IMMEDIATE: case ATA_SLEEP: ahci_write_fis_d2h(p, slot, cfis, ATA_S_READY | ATA_S_DSC); break; case ATA_ATAPI_IDENTIFY: handle_atapi_identify(p, slot, cfis); break; case ATA_PACKET_CMD: if (!p->atapi) { p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR; p->is |= AHCI_P_IX_TFE; ahci_generate_intr(p->pr_sc); } else handle_packet_cmd(p, slot, cfis); break; default: WPRINTF("Unsupported cmd:%02x\n", cfis[2]); p->tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR; p->is |= AHCI_P_IX_TFE; ahci_generate_intr(p->pr_sc); break; } } static void ahci_handle_slot(struct ahci_port *p, int slot) { struct ahci_cmd_hdr *hdr; struct ahci_prdt_entry *prdt; struct pci_ahci_softc *sc; uint8_t *cfis; int cfl; sc = p->pr_sc; hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE); cfl = (hdr->flags & 0x1f) * 4; cfis = paddr_guest2host(ahci_ctx(sc), hdr->ctba, 0x80 + hdr->prdtl * sizeof(struct ahci_prdt_entry)); prdt = (struct ahci_prdt_entry *)(cfis + 0x80); #ifdef AHCI_DEBUG DPRINTF("\ncfis:"); for (i = 0; i < cfl; i++) { if (i % 10 == 0) DPRINTF("\n"); DPRINTF("%02x ", cfis[i]); } DPRINTF("\n"); for (i = 0; i < hdr->prdtl; i++) { DPRINTF("%d@%08"PRIx64"\n", prdt->dbc & 0x3fffff, prdt->dba); prdt++; } #endif if (cfis[0] != FIS_TYPE_REGH2D) { WPRINTF("Not a H2D FIS:%02x\n", cfis[0]); return; } if (cfis[1] & 0x80) { ahci_handle_cmd(p, slot, cfis); } else { if (cfis[15] & (1 << 2)) p->reset = 1; else if (p->reset) { p->reset = 0; ahci_port_reset(p); } p->ci &= ~(1 << slot); } } static void ahci_handle_port(struct ahci_port *p) { int i; if (!(p->cmd & AHCI_P_CMD_ST)) return; /* * Search for any new commands to issue ignoring those that * are already in-flight. */ for (i = 0; (i < 32) && p->ci; i++) { if ((p->ci & (1 << i)) && !(p->pending & (1 << i))) { p->cmd &= ~AHCI_P_CMD_CCS_MASK; p->cmd |= i << AHCI_P_CMD_CCS_SHIFT; ahci_handle_slot(p, i); } } } /* * blockif callback routine - this runs in the context of the blockif * i/o thread, so the mutex needs to be acquired. */ static void ata_ioreq_cb(struct blockif_req *br, int err) { struct ahci_cmd_hdr *hdr; struct ahci_ioreq *aior; struct ahci_port *p; struct pci_ahci_softc *sc; uint32_t tfd; uint8_t *cfis; int pending, slot, ncq; DPRINTF("%s %d\n", __func__, err); ncq = 0; aior = br->br_param; p = aior->io_pr; cfis = aior->cfis; slot = aior->slot; pending = aior->prdtl; sc = p->pr_sc; hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + slot * AHCI_CL_SIZE); if (cfis[2] == ATA_WRITE_FPDMA_QUEUED || cfis[2] == ATA_READ_FPDMA_QUEUED) ncq = 1; pthread_mutex_lock(&sc->mtx); /* * Delete the blockif request from the busy list */ TAILQ_REMOVE(&p->iobhd, aior, io_blist); /* * Move the blockif request back to the free list */ STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist); if (pending && !err) { ahci_handle_dma(p, slot, cfis, aior->done, hdr->prdtl - pending); goto out; } if (!err && aior->done == aior->len) { tfd = ATA_S_READY | ATA_S_DSC; if (ncq) hdr->prdbc = 0; else hdr->prdbc = aior->len; } else { tfd = (ATA_E_ABORT << 8) | ATA_S_READY | ATA_S_ERROR; hdr->prdbc = 0; if (ncq) p->serr |= (1 << slot); } if (ncq) { p->sact &= ~(1 << slot); ahci_write_fis_sdb(p, slot, tfd); } else ahci_write_fis_d2h(p, slot, cfis, tfd); /* * This command is now complete. */ p->pending &= ~(1 << slot); ahci_check_stopped(p); out: pthread_mutex_unlock(&sc->mtx); DPRINTF("%s exit\n", __func__); } static void atapi_ioreq_cb(struct blockif_req *br, int err) { struct ahci_cmd_hdr *hdr; struct ahci_ioreq *aior; struct ahci_port *p; struct pci_ahci_softc *sc; uint8_t *cfis; uint32_t tfd; int pending, slot; DPRINTF("%s %d\n", __func__, err); aior = br->br_param; p = aior->io_pr; cfis = aior->cfis; slot = aior->slot; pending = aior->prdtl; sc = p->pr_sc; hdr = (struct ahci_cmd_hdr *)(p->cmd_lst + aior->slot * AHCI_CL_SIZE); pthread_mutex_lock(&sc->mtx); /* * Delete the blockif request from the busy list */ TAILQ_REMOVE(&p->iobhd, aior, io_blist); /* * Move the blockif request back to the free list */ STAILQ_INSERT_TAIL(&p->iofhd, aior, io_flist); if (pending && !err) { atapi_read(p, slot, cfis, aior->done, hdr->prdtl - pending); goto out; } if (!err && aior->done == aior->len) { tfd = ATA_S_READY | ATA_S_DSC; hdr->prdbc = aior->len; } else { p->sense_key = ATA_SENSE_ILLEGAL_REQUEST; p->asc = 0x21; tfd = (p->sense_key << 12) | ATA_S_READY | ATA_S_ERROR; hdr->prdbc = 0; } cfis[4] = (cfis[4] & ~7) | ATA_I_CMD | ATA_I_IN; ahci_write_fis_d2h(p, slot, cfis, tfd); /* * This command is now complete. */ p->pending &= ~(1 << slot); ahci_check_stopped(p); out: pthread_mutex_unlock(&sc->mtx); DPRINTF("%s exit\n", __func__); } static void pci_ahci_ioreq_init(struct ahci_port *pr) { struct ahci_ioreq *vr; int i; pr->ioqsz = blockif_queuesz(pr->bctx); pr->ioreq = calloc(pr->ioqsz, sizeof(struct ahci_ioreq)); STAILQ_INIT(&pr->iofhd); /* * Add all i/o request entries to the free queue */ for (i = 0; i < pr->ioqsz; i++) { vr = &pr->ioreq[i]; vr->io_pr = pr; if (!pr->atapi) vr->io_req.br_callback = ata_ioreq_cb; else vr->io_req.br_callback = atapi_ioreq_cb; vr->io_req.br_param = vr; STAILQ_INSERT_TAIL(&pr->iofhd, vr, io_flist); } TAILQ_INIT(&pr->iobhd); } static void pci_ahci_port_write(struct pci_ahci_softc *sc, uint64_t offset, uint64_t value) { int port = (offset - AHCI_OFFSET) / AHCI_STEP; offset = (offset - AHCI_OFFSET) % AHCI_STEP; struct ahci_port *p = &sc->port[port]; DPRINTF("pci_ahci_port %d: write offset 0x%"PRIx64" value 0x%"PRIx64"\n", port, offset, value); switch (offset) { case AHCI_P_CLB: p->clb = value; break; case AHCI_P_CLBU: p->clbu = value; break; case AHCI_P_FB: p->fb = value; break; case AHCI_P_FBU: p->fbu = value; break; case AHCI_P_IS: p->is &= ~value; break; case AHCI_P_IE: p->ie = value & 0xFDC000FF; ahci_generate_intr(sc); break; case AHCI_P_CMD: { p->cmd = value; if (!(value & AHCI_P_CMD_ST)) { ahci_port_stop(p); } else { uint64_t clb; p->cmd |= AHCI_P_CMD_CR; clb = (uint64_t)p->clbu << 32 | p->clb; p->cmd_lst = paddr_guest2host(ahci_ctx(sc), clb, AHCI_CL_SIZE * AHCI_MAX_SLOTS); } if (value & AHCI_P_CMD_FRE) { uint64_t fb; p->cmd |= AHCI_P_CMD_FR; fb = (uint64_t)p->fbu << 32 | p->fb; /* we don't support FBSCP, so rfis size is 256Bytes */ p->rfis = paddr_guest2host(ahci_ctx(sc), fb, 256); } else { p->cmd &= ~AHCI_P_CMD_FR; } if (value & AHCI_P_CMD_CLO) { p->tfd = 0; p->cmd &= ~AHCI_P_CMD_CLO; } ahci_handle_port(p); break; } case AHCI_P_TFD: case AHCI_P_SIG: case AHCI_P_SSTS: WPRINTF("pci_ahci_port: read only registers 0x%"PRIx64"\n", offset); break; case AHCI_P_SCTL: if (!(p->cmd & AHCI_P_CMD_ST)) { if (value & ATA_SC_DET_RESET) ahci_port_reset(p); p->sctl = value; } break; case AHCI_P_SERR: p->serr &= ~value; break; case AHCI_P_SACT: p->sact |= value; break; case AHCI_P_CI: p->ci |= value; ahci_handle_port(p); break; case AHCI_P_SNTF: case AHCI_P_FBS: default: break; } } static void pci_ahci_host_write(struct pci_ahci_softc *sc, uint64_t offset, uint64_t value) { DPRINTF("pci_ahci_host: write offset 0x%"PRIx64" value 0x%"PRIx64"\n", offset, value); switch (offset) { case AHCI_CAP: case AHCI_PI: case AHCI_VS: case AHCI_CAP2: DPRINTF("pci_ahci_host: read only registers 0x%"PRIx64"\n", offset); break; case AHCI_GHC: if (value & AHCI_GHC_HR) ahci_reset(sc); else if (value & AHCI_GHC_IE) { sc->ghc |= AHCI_GHC_IE; ahci_generate_intr(sc); } break; case AHCI_IS: sc->is &= ~value; ahci_generate_intr(sc); break; default: break; } } static void pci_ahci_write(struct vmctx *ctx, int vcpu, struct pci_devinst *pi, int baridx, uint64_t offset, int size, uint64_t value) { struct pci_ahci_softc *sc = pi->pi_arg; assert(baridx == 5); assert(size == 4); pthread_mutex_lock(&sc->mtx); if (offset < AHCI_OFFSET) pci_ahci_host_write(sc, offset, value); else if (offset < AHCI_OFFSET + sc->ports * AHCI_STEP) pci_ahci_port_write(sc, offset, value); else WPRINTF("pci_ahci: unknown i/o write offset 0x%"PRIx64"\n", offset); pthread_mutex_unlock(&sc->mtx); } static uint64_t pci_ahci_host_read(struct pci_ahci_softc *sc, uint64_t offset) { uint32_t value; switch (offset) { case AHCI_CAP: case AHCI_GHC: case AHCI_IS: case AHCI_PI: case AHCI_VS: case AHCI_CCCC: case AHCI_CCCP: case AHCI_EM_LOC: case AHCI_EM_CTL: case AHCI_CAP2: { uint32_t *p = &sc->cap; p += (offset - AHCI_CAP) / sizeof(uint32_t); value = *p; break; } default: value = 0; break; } DPRINTF("pci_ahci_host: read offset 0x%"PRIx64" value 0x%x\n", offset, value); return (value); } static uint64_t pci_ahci_port_read(struct pci_ahci_softc *sc, uint64_t offset) { uint32_t value; int port = (offset - AHCI_OFFSET) / AHCI_STEP; offset = (offset - AHCI_OFFSET) % AHCI_STEP; switch (offset) { case AHCI_P_CLB: case AHCI_P_CLBU: case AHCI_P_FB: case AHCI_P_FBU: case AHCI_P_IS: case AHCI_P_IE: case AHCI_P_CMD: case AHCI_P_TFD: case AHCI_P_SIG: case AHCI_P_SSTS: case AHCI_P_SCTL: case AHCI_P_SERR: case AHCI_P_SACT: case AHCI_P_CI: case AHCI_P_SNTF: case AHCI_P_FBS: { uint32_t *p= &sc->port[port].clb; p += (offset - AHCI_P_CLB) / sizeof(uint32_t); value = *p; break; } default: value = 0; break; } DPRINTF("pci_ahci_port %d: read offset 0x%"PRIx64" value 0x%x\n", port, offset, value); return value; } static uint64_t pci_ahci_read(struct vmctx *ctx, int vcpu, struct pci_devinst *pi, int baridx, uint64_t offset, int size) { struct pci_ahci_softc *sc = pi->pi_arg; uint32_t value; assert(baridx == 5); assert(size == 4); pthread_mutex_lock(&sc->mtx); if (offset < AHCI_OFFSET) value = pci_ahci_host_read(sc, offset); else if (offset < AHCI_OFFSET + sc->ports * AHCI_STEP) value = pci_ahci_port_read(sc, offset); else { value = 0; WPRINTF("pci_ahci: unknown i/o read offset 0x%"PRIx64"\n", offset); } pthread_mutex_unlock(&sc->mtx); return (value); } static int pci_ahci_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts, int atapi) { char bident[sizeof("XX:X:X")]; struct blockif_ctxt *bctxt; struct pci_ahci_softc *sc; int ret, slots; ret = 0; if (opts == NULL) { fprintf(stderr, "pci_ahci: backing device required\n"); return (1); } #ifdef AHCI_DEBUG dbg = fopen("/tmp/log", "w+"); #endif sc = calloc(1, sizeof(struct pci_ahci_softc)); pi->pi_arg = sc; sc->asc_pi = pi; sc->ports = MAX_PORTS; /* * Only use port 0 for a backing device. All other ports will be * marked as unused */ sc->port[0].atapi = atapi; /* * Attempt to open the backing image. Use the PCI * slot/func for the identifier string. */ snprintf(bident, sizeof(bident), "%d:%d", pi->pi_slot, pi->pi_func); bctxt = blockif_open(opts, bident); if (bctxt == NULL) { ret = 1; goto open_fail; } sc->port[0].bctx = bctxt; sc->port[0].pr_sc = sc; /* * Allocate blockif request structures and add them * to the free list */ pci_ahci_ioreq_init(&sc->port[0]); pthread_mutex_init(&sc->mtx, NULL); /* Intel ICH8 AHCI */ slots = sc->port[0].ioqsz; if (slots > 32) slots = 32; --slots; sc->cap = AHCI_CAP_64BIT | AHCI_CAP_SNCQ | AHCI_CAP_SSNTF | AHCI_CAP_SMPS | AHCI_CAP_SSS | AHCI_CAP_SALP | AHCI_CAP_SAL | AHCI_CAP_SCLO | (0x3 << AHCI_CAP_ISS_SHIFT)| AHCI_CAP_PMD | AHCI_CAP_SSC | AHCI_CAP_PSC | (slots << AHCI_CAP_NCS_SHIFT) | AHCI_CAP_SXS | (sc->ports - 1); /* Only port 0 implemented */ sc->pi = 1; sc->vs = 0x10300; sc->cap2 = AHCI_CAP2_APST; ahci_reset(sc); pci_set_cfgdata16(pi, PCIR_DEVICE, 0x2821); pci_set_cfgdata16(pi, PCIR_VENDOR, 0x8086); pci_set_cfgdata8(pi, PCIR_CLASS, PCIC_STORAGE); pci_set_cfgdata8(pi, PCIR_SUBCLASS, PCIS_STORAGE_SATA); pci_set_cfgdata8(pi, PCIR_PROGIF, PCIP_STORAGE_SATA_AHCI_1_0); pci_emul_add_msicap(pi, 1); pci_emul_alloc_bar(pi, 5, PCIBAR_MEM32, AHCI_OFFSET + sc->ports * AHCI_STEP); pci_lintr_request(pi); open_fail: if (ret) { - blockif_close(sc->port[0].bctx); + if (sc->port[0].bctx != NULL) + blockif_close(sc->port[0].bctx); free(sc); } return (ret); } static int pci_ahci_hd_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) { return (pci_ahci_init(ctx, pi, opts, 0)); } static int pci_ahci_atapi_init(struct vmctx *ctx, struct pci_devinst *pi, char *opts) { return (pci_ahci_init(ctx, pi, opts, 1)); } /* * Use separate emulation names to distinguish drive and atapi devices */ struct pci_devemu pci_de_ahci_hd = { .pe_emu = "ahci-hd", .pe_init = pci_ahci_hd_init, .pe_barwrite = pci_ahci_write, .pe_barread = pci_ahci_read }; PCI_EMUL_SET(pci_de_ahci_hd); struct pci_devemu pci_de_ahci_cd = { .pe_emu = "ahci-cd", .pe_init = pci_ahci_atapi_init, .pe_barwrite = pci_ahci_write, .pe_barread = pci_ahci_read }; PCI_EMUL_SET(pci_de_ahci_cd); Index: user/marcel/libvdsk/bhyve/rtc.c =================================================================== --- user/marcel/libvdsk/bhyve/rtc.c (revision 279574) +++ user/marcel/libvdsk/bhyve/rtc.c (revision 279575) @@ -1,382 +1,129 @@ /*- * Copyright (c) 2011 NetApp, 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 NETAPP, INC ``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 NETAPP, INC 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. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include -#include -#include -#include #include #include #include #include #include "acpi.h" -#include "inout.h" #include "pci_lpc.h" #include "rtc.h" -#define IO_RTC 0x70 +#define IO_RTC 0x70 -#define RTC_SEC 0x00 /* seconds */ -#define RTC_SEC_ALARM 0x01 -#define RTC_MIN 0x02 -#define RTC_MIN_ALARM 0x03 -#define RTC_HRS 0x04 -#define RTC_HRS_ALARM 0x05 -#define RTC_WDAY 0x06 -#define RTC_DAY 0x07 -#define RTC_MONTH 0x08 -#define RTC_YEAR 0x09 -#define RTC_CENTURY 0x32 /* current century */ - -#define RTC_STATUSA 0xA -#define RTCSA_TUP 0x80 /* time update, don't look now */ - -#define RTC_STATUSB 0xB -#define RTCSB_DST 0x01 -#define RTCSB_24HR 0x02 -#define RTCSB_BIN 0x04 /* 0 = BCD, 1 = Binary */ -#define RTCSB_PINTR 0x40 /* 1 = enable periodic clock interrupt */ -#define RTCSB_HALT 0x80 /* stop clock updates */ - -#define RTC_INTR 0x0c /* status register C (R) interrupt source */ - -#define RTC_STATUSD 0x0d /* status register D (R) Lost Power */ -#define RTCSD_PWR 0x80 /* clock power OK */ - -#define RTC_NVRAM_START 0x0e -#define RTC_NVRAM_END 0x7f -#define RTC_NVRAM_SZ (128 - RTC_NVRAM_START) -#define nvoff(x) ((x) - RTC_NVRAM_START) - -#define RTC_DIAG 0x0e -#define RTC_RSTCODE 0x0f -#define RTC_EQUIPMENT 0x14 #define RTC_LMEM_LSB 0x34 #define RTC_LMEM_MSB 0x35 #define RTC_HMEM_LSB 0x5b #define RTC_HMEM_SB 0x5c #define RTC_HMEM_MSB 0x5d #define m_64KB (64*1024) #define m_16MB (16*1024*1024) #define m_4GB (4ULL*1024*1024*1024) -static int addr; - -static uint8_t rtc_nvram[RTC_NVRAM_SZ]; - -/* XXX initialize these to default values as they would be from BIOS */ -static uint8_t status_a, status_b; - -static struct { - uint8_t hours; - uint8_t mins; - uint8_t secs; -} rtc_alarm; - -static u_char const bin2bcd_data[] = { - 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07, 0x08, 0x09, - 0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, - 0x20, 0x21, 0x22, 0x23, 0x24, 0x25, 0x26, 0x27, 0x28, 0x29, - 0x30, 0x31, 0x32, 0x33, 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, - 0x40, 0x41, 0x42, 0x43, 0x44, 0x45, 0x46, 0x47, 0x48, 0x49, - 0x50, 0x51, 0x52, 0x53, 0x54, 0x55, 0x56, 0x57, 0x58, 0x59, - 0x60, 0x61, 0x62, 0x63, 0x64, 0x65, 0x66, 0x67, 0x68, 0x69, - 0x70, 0x71, 0x72, 0x73, 0x74, 0x75, 0x76, 0x77, 0x78, 0x79, - 0x80, 0x81, 0x82, 0x83, 0x84, 0x85, 0x86, 0x87, 0x88, 0x89, - 0x90, 0x91, 0x92, 0x93, 0x94, 0x95, 0x96, 0x97, 0x98, 0x99 -}; -#define bin2bcd(bin) (bin2bcd_data[bin]) - -#define rtcout(val) ((status_b & RTCSB_BIN) ? (val) : bin2bcd((val))) - -static void -timevalfix(struct timeval *t1) +/* + * Returns the current RTC time as number of seconds since 00:00:00 Jan 1, 1970 + */ +static time_t +rtc_time(struct vmctx *ctx, int use_localtime) { - - if (t1->tv_usec < 0) { - t1->tv_sec--; - t1->tv_usec += 1000000; - } - if (t1->tv_usec >= 1000000) { - t1->tv_sec++; - t1->tv_usec -= 1000000; - } -} - -static void -timevalsub(struct timeval *t1, const struct timeval *t2) -{ - - t1->tv_sec -= t2->tv_sec; - t1->tv_usec -= t2->tv_usec; - timevalfix(t1); -} - -static int -rtc_addr_handler(struct vmctx *ctx, int vcpu, int in, int port, int bytes, - uint32_t *eax, void *arg) -{ - if (bytes != 1) - return (-1); - - if (in) { - /* straight read of this register will return 0xFF */ - *eax = 0xff; - return (0); - } - - switch (*eax & 0x7f) { - case RTC_SEC: - case RTC_SEC_ALARM: - case RTC_MIN: - case RTC_MIN_ALARM: - case RTC_HRS: - case RTC_HRS_ALARM: - case RTC_WDAY: - case RTC_DAY: - case RTC_MONTH: - case RTC_YEAR: - case RTC_STATUSA: - case RTC_STATUSB: - case RTC_INTR: - case RTC_STATUSD: - case RTC_NVRAM_START ... RTC_NVRAM_END: - break; - default: - return (-1); - } - - addr = *eax & 0x7f; - return (0); -} - -static int -rtc_data_handler(struct vmctx *ctx, int vcpu, int in, int port, int bytes, - uint32_t *eax, void *arg) -{ - int hour; + struct tm tm; time_t t; - struct timeval cur, delta; - static struct timeval last; - static struct tm tm; - - if (bytes != 1) - return (-1); - - gettimeofday(&cur, NULL); - - /* - * Increment the cached time only once per second so we can guarantee - * that the guest has at least one second to read the hour:min:sec - * separately and still get a coherent view of the time. - */ - delta = cur; - timevalsub(&delta, &last); - if (delta.tv_sec >= 1 && (status_b & RTCSB_HALT) == 0) { - t = cur.tv_sec; + time(&t); + if (use_localtime) { localtime_r(&t, &tm); - last = cur; + t = timegm(&tm); } - - if (in) { - switch (addr) { - case RTC_SEC_ALARM: - *eax = rtc_alarm.secs; - break; - case RTC_MIN_ALARM: - *eax = rtc_alarm.mins; - break; - case RTC_HRS_ALARM: - *eax = rtc_alarm.hours; - break; - case RTC_SEC: - *eax = rtcout(tm.tm_sec); - return (0); - case RTC_MIN: - *eax = rtcout(tm.tm_min); - return (0); - case RTC_HRS: - if (status_b & RTCSB_24HR) - hour = tm.tm_hour; - else - hour = (tm.tm_hour % 12) + 1; - - *eax = rtcout(hour); - - /* - * If we are representing time in the 12-hour format - * then set the MSB to indicate PM. - */ - if ((status_b & RTCSB_24HR) == 0 && tm.tm_hour >= 12) - *eax |= 0x80; - - return (0); - case RTC_WDAY: - *eax = rtcout(tm.tm_wday + 1); - return (0); - case RTC_DAY: - *eax = rtcout(tm.tm_mday); - return (0); - case RTC_MONTH: - *eax = rtcout(tm.tm_mon + 1); - return (0); - case RTC_YEAR: - *eax = rtcout(tm.tm_year % 100); - return (0); - case RTC_STATUSA: - *eax = status_a; - return (0); - case RTC_STATUSB: - *eax = status_b; - return (0); - case RTC_INTR: - *eax = 0; - return (0); - case RTC_STATUSD: - *eax = RTCSD_PWR; - return (0); - case RTC_NVRAM_START ... RTC_NVRAM_END: - *eax = rtc_nvram[addr - RTC_NVRAM_START]; - return (0); - default: - return (-1); - } - } - - switch (addr) { - case RTC_STATUSA: - status_a = *eax & ~RTCSA_TUP; - break; - case RTC_STATUSB: - /* XXX not implemented yet XXX */ - if (*eax & RTCSB_PINTR) - return (-1); - status_b = *eax; - break; - case RTC_STATUSD: - /* ignore write */ - break; - case RTC_SEC_ALARM: - rtc_alarm.secs = *eax; - break; - case RTC_MIN_ALARM: - rtc_alarm.mins = *eax; - break; - case RTC_HRS_ALARM: - rtc_alarm.hours = *eax; - break; - case RTC_SEC: - case RTC_MIN: - case RTC_HRS: - case RTC_WDAY: - case RTC_DAY: - case RTC_MONTH: - case RTC_YEAR: - /* - * Ignore writes to the time of day registers - */ - break; - case RTC_NVRAM_START ... RTC_NVRAM_END: - rtc_nvram[addr - RTC_NVRAM_START] = *eax; - break; - default: - return (-1); - } - return (0); + return (t); } void -rtc_init(struct vmctx *ctx) +rtc_init(struct vmctx *ctx, int use_localtime) { - struct timeval cur; - struct tm tm; size_t himem; size_t lomem; int err; - err = gettimeofday(&cur, NULL); - assert(err == 0); - (void) localtime_r(&cur.tv_sec, &tm); - - memset(rtc_nvram, 0, sizeof(rtc_nvram)); - - rtc_nvram[nvoff(RTC_CENTURY)] = bin2bcd((tm.tm_year + 1900) / 100); - /* XXX init diag/reset code/equipment/checksum ? */ /* * Report guest memory size in nvram cells as required by UEFI. * Little-endian encoding. * 0x34/0x35 - 64KB chunks above 16MB, below 4GB * 0x5b/0x5c/0x5d - 64KB chunks above 4GB */ lomem = (vm_get_lowmem_size(ctx) - m_16MB) / m_64KB; - rtc_nvram[nvoff(RTC_LMEM_LSB)] = lomem; - rtc_nvram[nvoff(RTC_LMEM_MSB)] = lomem >> 8; + err = vm_rtc_write(ctx, RTC_LMEM_LSB, lomem); + assert(err == 0); + err = vm_rtc_write(ctx, RTC_LMEM_MSB, lomem >> 8); + assert(err == 0); himem = vm_get_highmem_size(ctx) / m_64KB; - rtc_nvram[nvoff(RTC_HMEM_LSB)] = himem; - rtc_nvram[nvoff(RTC_HMEM_SB)] = himem >> 8; - rtc_nvram[nvoff(RTC_HMEM_MSB)] = himem >> 16; -} + err = vm_rtc_write(ctx, RTC_HMEM_LSB, himem); + assert(err == 0); + err = vm_rtc_write(ctx, RTC_HMEM_SB, himem >> 8); + assert(err == 0); + err = vm_rtc_write(ctx, RTC_HMEM_MSB, himem >> 16); + assert(err == 0); -INOUT_PORT(rtc, IO_RTC, IOPORT_F_INOUT, rtc_addr_handler); -INOUT_PORT(rtc, IO_RTC + 1, IOPORT_F_INOUT, rtc_data_handler); + err = vm_rtc_settime(ctx, rtc_time(ctx, use_localtime)); + assert(err == 0); +} static void rtc_dsdt(void) { dsdt_line(""); dsdt_line("Device (RTC)"); dsdt_line("{"); dsdt_line(" Name (_HID, EisaId (\"PNP0B00\"))"); dsdt_line(" Name (_CRS, ResourceTemplate ()"); dsdt_line(" {"); dsdt_indent(2); dsdt_fixed_ioport(IO_RTC, 2); dsdt_fixed_irq(8); dsdt_unindent(2); dsdt_line(" })"); dsdt_line("}"); } LPC_DSDT(rtc_dsdt); /* * Reserve the extended RTC I/O ports although they are not emulated at this * time. */ SYSRES_IO(0x72, 6); Index: user/marcel/libvdsk/bhyve/rtc.h =================================================================== --- user/marcel/libvdsk/bhyve/rtc.h (revision 279574) +++ user/marcel/libvdsk/bhyve/rtc.h (revision 279575) @@ -1,34 +1,34 @@ /* * Copyright (c) 2013 Peter Grehan * 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 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. * * $FreeBSD$ */ #ifndef _RTC_H_ #define _RTC_H_ -void rtc_init(struct vmctx *ctx); +void rtc_init(struct vmctx *ctx, int use_localtime); #endif /* _RTC_H_ */ Index: user/marcel/libvdsk/bhyve/task_switch.c =================================================================== --- user/marcel/libvdsk/bhyve/task_switch.c (revision 279574) +++ user/marcel/libvdsk/bhyve/task_switch.c (revision 279575) @@ -1,946 +1,936 @@ /*- * Copyright (c) 2014 Neel Natu * 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 AUTHOR ``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 __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "bhyverun.h" /* * Using 'struct i386tss' is tempting but causes myriad sign extension * issues because all of its fields are defined as signed integers. */ struct tss32 { uint16_t tss_link; uint16_t rsvd1; uint32_t tss_esp0; uint16_t tss_ss0; uint16_t rsvd2; uint32_t tss_esp1; uint16_t tss_ss1; uint16_t rsvd3; uint32_t tss_esp2; uint16_t tss_ss2; uint16_t rsvd4; uint32_t tss_cr3; uint32_t tss_eip; uint32_t tss_eflags; uint32_t tss_eax; uint32_t tss_ecx; uint32_t tss_edx; uint32_t tss_ebx; uint32_t tss_esp; uint32_t tss_ebp; uint32_t tss_esi; uint32_t tss_edi; uint16_t tss_es; uint16_t rsvd5; uint16_t tss_cs; uint16_t rsvd6; uint16_t tss_ss; uint16_t rsvd7; uint16_t tss_ds; uint16_t rsvd8; uint16_t tss_fs; uint16_t rsvd9; uint16_t tss_gs; uint16_t rsvd10; uint16_t tss_ldt; uint16_t rsvd11; uint16_t tss_trap; uint16_t tss_iomap; }; CTASSERT(sizeof(struct tss32) == 104); #define SEL_START(sel) (((sel) & ~0x7)) #define SEL_LIMIT(sel) (((sel) | 0x7)) #define TSS_BUSY(type) (((type) & 0x2) != 0) static uint64_t GETREG(struct vmctx *ctx, int vcpu, int reg) { uint64_t val; int error; error = vm_get_register(ctx, vcpu, reg, &val); assert(error == 0); return (val); } static void SETREG(struct vmctx *ctx, int vcpu, int reg, uint64_t val) { int error; error = vm_set_register(ctx, vcpu, reg, val); assert(error == 0); } static struct seg_desc usd_to_seg_desc(struct user_segment_descriptor *usd) { struct seg_desc seg_desc; seg_desc.base = (u_int)USD_GETBASE(usd); if (usd->sd_gran) seg_desc.limit = (u_int)(USD_GETLIMIT(usd) << 12) | 0xfff; else seg_desc.limit = (u_int)USD_GETLIMIT(usd); seg_desc.access = usd->sd_type | usd->sd_dpl << 5 | usd->sd_p << 7; seg_desc.access |= usd->sd_xx << 12; seg_desc.access |= usd->sd_def32 << 14; seg_desc.access |= usd->sd_gran << 15; return (seg_desc); } /* * Inject an exception with an error code that is a segment selector. * The format of the error code is described in section 6.13, "Error Code", * Intel SDM volume 3. * * Bit 0 (EXT) denotes whether the exception occurred during delivery * of an external event like an interrupt. * * Bit 1 (IDT) indicates whether the selector points to a gate descriptor * in the IDT. * * Bit 2(GDT/LDT) has the usual interpretation of Table Indicator (TI). */ static void sel_exception(struct vmctx *ctx, int vcpu, int vector, uint16_t sel, int ext) { /* * Bit 2 from the selector is retained as-is in the error code. * * Bit 1 can be safely cleared because none of the selectors * encountered during task switch emulation refer to a task * gate in the IDT. * * Bit 0 is set depending on the value of 'ext'. */ sel &= ~0x3; if (ext) sel |= 0x1; vm_inject_fault(ctx, vcpu, vector, 1, sel); } /* * Return 0 if the selector 'sel' in within the limits of the GDT/LDT * and non-zero otherwise. */ static int desc_table_limit_check(struct vmctx *ctx, int vcpu, uint16_t sel) { uint64_t base; uint32_t limit, access; int error, reg; reg = ISLDT(sel) ? VM_REG_GUEST_LDTR : VM_REG_GUEST_GDTR; error = vm_get_desc(ctx, vcpu, reg, &base, &limit, &access); assert(error == 0); if (reg == VM_REG_GUEST_LDTR) { if (SEG_DESC_UNUSABLE(access) || !SEG_DESC_PRESENT(access)) return (-1); } if (limit < SEL_LIMIT(sel)) return (-1); else return (0); } /* * Read/write the segment descriptor 'desc' into the GDT/LDT slot referenced * by the selector 'sel'. * * Returns 0 on success. * Returns 1 if an exception was injected into the guest. * Returns -1 otherwise. */ static int desc_table_rw(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, uint16_t sel, struct user_segment_descriptor *desc, bool doread) { struct iovec iov[2]; uint64_t base; uint32_t limit, access; int error, reg; reg = ISLDT(sel) ? VM_REG_GUEST_LDTR : VM_REG_GUEST_GDTR; error = vm_get_desc(ctx, vcpu, reg, &base, &limit, &access); assert(error == 0); assert(limit >= SEL_LIMIT(sel)); error = vm_copy_setup(ctx, vcpu, paging, base + SEL_START(sel), sizeof(*desc), doread ? PROT_READ : PROT_WRITE, iov, nitems(iov)); if (error == 0) { if (doread) vm_copyin(ctx, vcpu, iov, desc, sizeof(*desc)); else vm_copyout(ctx, vcpu, desc, iov, sizeof(*desc)); } return (error); } static int desc_table_read(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, uint16_t sel, struct user_segment_descriptor *desc) { return (desc_table_rw(ctx, vcpu, paging, sel, desc, true)); } static int desc_table_write(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, uint16_t sel, struct user_segment_descriptor *desc) { return (desc_table_rw(ctx, vcpu, paging, sel, desc, false)); } /* * Read the TSS descriptor referenced by 'sel' into 'desc'. * * Returns 0 on success. * Returns 1 if an exception was injected into the guest. * Returns -1 otherwise. */ static int read_tss_descriptor(struct vmctx *ctx, int vcpu, struct vm_task_switch *ts, uint16_t sel, struct user_segment_descriptor *desc) { struct vm_guest_paging sup_paging; int error; assert(!ISLDT(sel)); assert(IDXSEL(sel) != 0); /* Fetch the new TSS descriptor */ if (desc_table_limit_check(ctx, vcpu, sel)) { if (ts->reason == TSR_IRET) sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); else sel_exception(ctx, vcpu, IDT_GP, sel, ts->ext); return (1); } sup_paging = ts->paging; sup_paging.cpl = 0; /* implicit supervisor mode */ error = desc_table_read(ctx, vcpu, &sup_paging, sel, desc); return (error); } static bool code_desc(int sd_type) { /* code descriptor */ return ((sd_type & 0x18) == 0x18); } static bool stack_desc(int sd_type) { /* writable data descriptor */ return ((sd_type & 0x1A) == 0x12); } static bool data_desc(int sd_type) { /* data descriptor or a readable code descriptor */ return ((sd_type & 0x18) == 0x10 || (sd_type & 0x1A) == 0x1A); } static bool ldt_desc(int sd_type) { return (sd_type == SDT_SYSLDT); } /* * Validate the descriptor 'seg_desc' associated with 'segment'. * * Returns 0 on success. * Returns 1 if an exception was injected into the guest. * Returns -1 otherwise. */ static int validate_seg_desc(struct vmctx *ctx, int vcpu, struct vm_task_switch *ts, int segment, struct seg_desc *seg_desc) { struct vm_guest_paging sup_paging; struct user_segment_descriptor usd; int error, idtvec; int cpl, dpl, rpl; uint16_t sel, cs; bool ldtseg, codeseg, stackseg, dataseg, conforming; ldtseg = codeseg = stackseg = dataseg = false; switch (segment) { case VM_REG_GUEST_LDTR: ldtseg = true; break; case VM_REG_GUEST_CS: codeseg = true; break; case VM_REG_GUEST_SS: stackseg = true; break; case VM_REG_GUEST_DS: case VM_REG_GUEST_ES: case VM_REG_GUEST_FS: case VM_REG_GUEST_GS: dataseg = true; break; default: assert(0); } /* Get the segment selector */ sel = GETREG(ctx, vcpu, segment); /* LDT selector must point into the GDT */ if (ldtseg && ISLDT(sel)) { sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); return (1); } /* Descriptor table limit check */ if (desc_table_limit_check(ctx, vcpu, sel)) { sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); return (1); } /* NULL selector */ if (IDXSEL(sel) == 0) { /* Code and stack segment selectors cannot be NULL */ if (codeseg || stackseg) { sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); return (1); } seg_desc->base = 0; seg_desc->limit = 0; seg_desc->access = 0x10000; /* unusable */ return (0); } /* Read the descriptor from the GDT/LDT */ sup_paging = ts->paging; sup_paging.cpl = 0; /* implicit supervisor mode */ error = desc_table_read(ctx, vcpu, &sup_paging, sel, &usd); if (error) return (error); /* Verify that the descriptor type is compatible with the segment */ if ((ldtseg && !ldt_desc(usd.sd_type)) || (codeseg && !code_desc(usd.sd_type)) || (dataseg && !data_desc(usd.sd_type)) || (stackseg && !stack_desc(usd.sd_type))) { sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); return (1); } /* Segment must be marked present */ if (!usd.sd_p) { if (ldtseg) idtvec = IDT_TS; else if (stackseg) idtvec = IDT_SS; else idtvec = IDT_NP; sel_exception(ctx, vcpu, idtvec, sel, ts->ext); return (1); } cs = GETREG(ctx, vcpu, VM_REG_GUEST_CS); cpl = cs & SEL_RPL_MASK; rpl = sel & SEL_RPL_MASK; dpl = usd.sd_dpl; if (stackseg && (rpl != cpl || dpl != cpl)) { sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); return (1); } if (codeseg) { conforming = (usd.sd_type & 0x4) ? true : false; if ((conforming && (cpl < dpl)) || (!conforming && (cpl != dpl))) { sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); return (1); } } if (dataseg) { /* * A data segment is always non-conforming except when it's * descriptor is a readable, conforming code segment. */ if (code_desc(usd.sd_type) && (usd.sd_type & 0x4) != 0) conforming = true; else conforming = false; if (!conforming && (rpl > dpl || cpl > dpl)) { sel_exception(ctx, vcpu, IDT_TS, sel, ts->ext); return (1); } } *seg_desc = usd_to_seg_desc(&usd); return (0); } static void tss32_save(struct vmctx *ctx, int vcpu, struct vm_task_switch *task_switch, uint32_t eip, struct tss32 *tss, struct iovec *iov) { /* General purpose registers */ tss->tss_eax = GETREG(ctx, vcpu, VM_REG_GUEST_RAX); tss->tss_ecx = GETREG(ctx, vcpu, VM_REG_GUEST_RCX); tss->tss_edx = GETREG(ctx, vcpu, VM_REG_GUEST_RDX); tss->tss_ebx = GETREG(ctx, vcpu, VM_REG_GUEST_RBX); tss->tss_esp = GETREG(ctx, vcpu, VM_REG_GUEST_RSP); tss->tss_ebp = GETREG(ctx, vcpu, VM_REG_GUEST_RBP); tss->tss_esi = GETREG(ctx, vcpu, VM_REG_GUEST_RSI); tss->tss_edi = GETREG(ctx, vcpu, VM_REG_GUEST_RDI); /* Segment selectors */ tss->tss_es = GETREG(ctx, vcpu, VM_REG_GUEST_ES); tss->tss_cs = GETREG(ctx, vcpu, VM_REG_GUEST_CS); tss->tss_ss = GETREG(ctx, vcpu, VM_REG_GUEST_SS); tss->tss_ds = GETREG(ctx, vcpu, VM_REG_GUEST_DS); tss->tss_fs = GETREG(ctx, vcpu, VM_REG_GUEST_FS); tss->tss_gs = GETREG(ctx, vcpu, VM_REG_GUEST_GS); /* eflags and eip */ tss->tss_eflags = GETREG(ctx, vcpu, VM_REG_GUEST_RFLAGS); if (task_switch->reason == TSR_IRET) tss->tss_eflags &= ~PSL_NT; tss->tss_eip = eip; /* Copy updated old TSS into guest memory */ vm_copyout(ctx, vcpu, tss, iov, sizeof(struct tss32)); } static void update_seg_desc(struct vmctx *ctx, int vcpu, int reg, struct seg_desc *sd) { int error; error = vm_set_desc(ctx, vcpu, reg, sd->base, sd->limit, sd->access); assert(error == 0); } /* * Update the vcpu registers to reflect the state of the new task. * * Returns 0 on success. * Returns 1 if an exception was injected into the guest. * Returns -1 otherwise. */ static int tss32_restore(struct vmctx *ctx, int vcpu, struct vm_task_switch *ts, uint16_t ot_sel, struct tss32 *tss, struct iovec *iov) { struct seg_desc seg_desc, seg_desc2; uint64_t *pdpte, maxphyaddr, reserved; uint32_t eflags; int error, i; bool nested; nested = false; if (ts->reason != TSR_IRET && ts->reason != TSR_JMP) { tss->tss_link = ot_sel; nested = true; } eflags = tss->tss_eflags; if (nested) eflags |= PSL_NT; /* LDTR */ SETREG(ctx, vcpu, VM_REG_GUEST_LDTR, tss->tss_ldt); /* PBDR */ if (ts->paging.paging_mode != PAGING_MODE_FLAT) { if (ts->paging.paging_mode == PAGING_MODE_PAE) { /* * XXX Assuming 36-bit MAXPHYADDR. */ maxphyaddr = (1UL << 36) - 1; pdpte = paddr_guest2host(ctx, tss->tss_cr3 & ~0x1f, 32); for (i = 0; i < 4; i++) { /* Check reserved bits if the PDPTE is valid */ if (!(pdpte[i] & 0x1)) continue; /* * Bits 2:1, 8:5 and bits above the processor's * maximum physical address are reserved. */ reserved = ~maxphyaddr | 0x1E6; if (pdpte[i] & reserved) { vm_inject_gp(ctx, vcpu); return (1); } } SETREG(ctx, vcpu, VM_REG_GUEST_PDPTE0, pdpte[0]); SETREG(ctx, vcpu, VM_REG_GUEST_PDPTE1, pdpte[1]); SETREG(ctx, vcpu, VM_REG_GUEST_PDPTE2, pdpte[2]); SETREG(ctx, vcpu, VM_REG_GUEST_PDPTE3, pdpte[3]); } SETREG(ctx, vcpu, VM_REG_GUEST_CR3, tss->tss_cr3); ts->paging.cr3 = tss->tss_cr3; } /* eflags and eip */ SETREG(ctx, vcpu, VM_REG_GUEST_RFLAGS, eflags); SETREG(ctx, vcpu, VM_REG_GUEST_RIP, tss->tss_eip); /* General purpose registers */ SETREG(ctx, vcpu, VM_REG_GUEST_RAX, tss->tss_eax); SETREG(ctx, vcpu, VM_REG_GUEST_RCX, tss->tss_ecx); SETREG(ctx, vcpu, VM_REG_GUEST_RDX, tss->tss_edx); SETREG(ctx, vcpu, VM_REG_GUEST_RBX, tss->tss_ebx); SETREG(ctx, vcpu, VM_REG_GUEST_RSP, tss->tss_esp); SETREG(ctx, vcpu, VM_REG_GUEST_RBP, tss->tss_ebp); SETREG(ctx, vcpu, VM_REG_GUEST_RSI, tss->tss_esi); SETREG(ctx, vcpu, VM_REG_GUEST_RDI, tss->tss_edi); /* Segment selectors */ SETREG(ctx, vcpu, VM_REG_GUEST_ES, tss->tss_es); SETREG(ctx, vcpu, VM_REG_GUEST_CS, tss->tss_cs); SETREG(ctx, vcpu, VM_REG_GUEST_SS, tss->tss_ss); SETREG(ctx, vcpu, VM_REG_GUEST_DS, tss->tss_ds); SETREG(ctx, vcpu, VM_REG_GUEST_FS, tss->tss_fs); SETREG(ctx, vcpu, VM_REG_GUEST_GS, tss->tss_gs); /* * If this is a nested task then write out the new TSS to update * the previous link field. */ if (nested) vm_copyout(ctx, vcpu, tss, iov, sizeof(*tss)); /* Validate segment descriptors */ error = validate_seg_desc(ctx, vcpu, ts, VM_REG_GUEST_LDTR, &seg_desc); if (error) return (error); update_seg_desc(ctx, vcpu, VM_REG_GUEST_LDTR, &seg_desc); /* * Section "Checks on Guest Segment Registers", Intel SDM, Vol 3. * * The SS and CS attribute checks on VM-entry are inter-dependent so * we need to make sure that both segments are valid before updating * either of them. This ensures that the VMCS state can pass the * VM-entry checks so the guest can handle any exception injected * during task switch emulation. */ error = validate_seg_desc(ctx, vcpu, ts, VM_REG_GUEST_CS, &seg_desc); if (error) return (error); error = validate_seg_desc(ctx, vcpu, ts, VM_REG_GUEST_SS, &seg_desc2); if (error) return (error); update_seg_desc(ctx, vcpu, VM_REG_GUEST_CS, &seg_desc); update_seg_desc(ctx, vcpu, VM_REG_GUEST_SS, &seg_desc2); ts->paging.cpl = tss->tss_cs & SEL_RPL_MASK; error = validate_seg_desc(ctx, vcpu, ts, VM_REG_GUEST_DS, &seg_desc); if (error) return (error); update_seg_desc(ctx, vcpu, VM_REG_GUEST_DS, &seg_desc); error = validate_seg_desc(ctx, vcpu, ts, VM_REG_GUEST_ES, &seg_desc); if (error) return (error); update_seg_desc(ctx, vcpu, VM_REG_GUEST_ES, &seg_desc); error = validate_seg_desc(ctx, vcpu, ts, VM_REG_GUEST_FS, &seg_desc); if (error) return (error); update_seg_desc(ctx, vcpu, VM_REG_GUEST_FS, &seg_desc); error = validate_seg_desc(ctx, vcpu, ts, VM_REG_GUEST_GS, &seg_desc); if (error) return (error); update_seg_desc(ctx, vcpu, VM_REG_GUEST_GS, &seg_desc); return (0); } /* * Push an error code on the stack of the new task. This is needed if the * task switch was triggered by a hardware exception that causes an error * code to be saved (e.g. #PF). * * Returns 0 on success. * Returns 1 if an exception was injected into the guest. * Returns -1 otherwise. */ static int push_errcode(struct vmctx *ctx, int vcpu, struct vm_guest_paging *paging, int task_type, uint32_t errcode) { struct iovec iov[2]; struct seg_desc seg_desc; int stacksize, bytes, error; uint64_t gla, cr0, rflags; uint32_t esp; uint16_t stacksel; cr0 = GETREG(ctx, vcpu, VM_REG_GUEST_CR0); rflags = GETREG(ctx, vcpu, VM_REG_GUEST_RFLAGS); stacksel = GETREG(ctx, vcpu, VM_REG_GUEST_SS); error = vm_get_desc(ctx, vcpu, VM_REG_GUEST_SS, &seg_desc.base, &seg_desc.limit, &seg_desc.access); assert(error == 0); /* * Section "Error Code" in the Intel SDM vol 3: the error code is * pushed on the stack as a doubleword or word (depending on the * default interrupt, trap or task gate size). */ if (task_type == SDT_SYS386BSY || task_type == SDT_SYS386TSS) bytes = 4; else bytes = 2; /* * PUSH instruction from Intel SDM vol 2: the 'B' flag in the * stack-segment descriptor determines the size of the stack * pointer outside of 64-bit mode. */ if (SEG_DESC_DEF32(seg_desc.access)) stacksize = 4; else stacksize = 2; esp = GETREG(ctx, vcpu, VM_REG_GUEST_RSP); esp -= bytes; if (vie_calculate_gla(paging->cpu_mode, VM_REG_GUEST_SS, &seg_desc, esp, bytes, stacksize, PROT_WRITE, &gla)) { sel_exception(ctx, vcpu, IDT_SS, stacksel, 1); return (1); } if (vie_alignment_check(paging->cpl, bytes, cr0, rflags, gla)) { vm_inject_ac(ctx, vcpu, 1); return (1); } error = vm_copy_setup(ctx, vcpu, paging, gla, bytes, PROT_WRITE, iov, nitems(iov)); if (error) return (error); vm_copyout(ctx, vcpu, &errcode, iov, bytes); SETREG(ctx, vcpu, VM_REG_GUEST_RSP, esp); return (0); } /* * Evaluate return value from helper functions and potentially return to * the VM run loop. * 0: success * +1: an exception was injected into the guest vcpu * -1: unrecoverable/programming error */ #define CHKERR(x) \ do { \ assert(((x) == 0) || ((x) == 1) || ((x) == -1)); \ if ((x) == -1) \ return (VMEXIT_ABORT); \ else if ((x) == 1) \ return (VMEXIT_CONTINUE); \ } while (0) int vmexit_task_switch(struct vmctx *ctx, struct vm_exit *vmexit, int *pvcpu) { struct seg_desc nt; struct tss32 oldtss, newtss; struct vm_task_switch *task_switch; struct vm_guest_paging *paging, sup_paging; struct user_segment_descriptor nt_desc, ot_desc; struct iovec nt_iov[2], ot_iov[2]; uint64_t cr0, ot_base; uint32_t eip, ot_lim, access; int error, ext, minlimit, nt_type, ot_type, vcpu; enum task_switch_reason reason; uint16_t nt_sel, ot_sel; task_switch = &vmexit->u.task_switch; nt_sel = task_switch->tsssel; ext = vmexit->u.task_switch.ext; reason = vmexit->u.task_switch.reason; paging = &vmexit->u.task_switch.paging; vcpu = *pvcpu; assert(paging->cpu_mode == CPU_MODE_PROTECTED); /* - * Calculate the %eip to store in the old TSS before modifying the - * 'inst_length'. + * Calculate the instruction pointer to store in the old TSS. */ eip = vmexit->rip + vmexit->inst_length; /* - * Set the 'inst_length' to '0'. - * - * If an exception is triggered during emulation of the task switch - * then the exception handler should return to the instruction that - * caused the task switch as opposed to the subsequent instruction. - */ - vmexit->inst_length = 0; - - /* * Section 4.6, "Access Rights" in Intel SDM Vol 3. * The following page table accesses are implicitly supervisor mode: * - accesses to GDT or LDT to load segment descriptors * - accesses to the task state segment during task switch */ sup_paging = *paging; sup_paging.cpl = 0; /* implicit supervisor mode */ /* Fetch the new TSS descriptor */ error = read_tss_descriptor(ctx, vcpu, task_switch, nt_sel, &nt_desc); CHKERR(error); nt = usd_to_seg_desc(&nt_desc); /* Verify the type of the new TSS */ nt_type = SEG_DESC_TYPE(nt.access); if (nt_type != SDT_SYS386BSY && nt_type != SDT_SYS386TSS && nt_type != SDT_SYS286BSY && nt_type != SDT_SYS286TSS) { sel_exception(ctx, vcpu, IDT_TS, nt_sel, ext); goto done; } /* TSS descriptor must have present bit set */ if (!SEG_DESC_PRESENT(nt.access)) { sel_exception(ctx, vcpu, IDT_NP, nt_sel, ext); goto done; } /* * TSS must have a minimum length of 104 bytes for a 32-bit TSS and * 44 bytes for a 16-bit TSS. */ if (nt_type == SDT_SYS386BSY || nt_type == SDT_SYS386TSS) minlimit = 104 - 1; else if (nt_type == SDT_SYS286BSY || nt_type == SDT_SYS286TSS) minlimit = 44 - 1; else minlimit = 0; assert(minlimit > 0); if (nt.limit < minlimit) { sel_exception(ctx, vcpu, IDT_TS, nt_sel, ext); goto done; } /* TSS must be busy if task switch is due to IRET */ if (reason == TSR_IRET && !TSS_BUSY(nt_type)) { sel_exception(ctx, vcpu, IDT_TS, nt_sel, ext); goto done; } /* * TSS must be available (not busy) if task switch reason is * CALL, JMP, exception or interrupt. */ if (reason != TSR_IRET && TSS_BUSY(nt_type)) { sel_exception(ctx, vcpu, IDT_GP, nt_sel, ext); goto done; } /* Fetch the new TSS */ error = vm_copy_setup(ctx, vcpu, &sup_paging, nt.base, minlimit + 1, PROT_READ | PROT_WRITE, nt_iov, nitems(nt_iov)); CHKERR(error); vm_copyin(ctx, vcpu, nt_iov, &newtss, minlimit + 1); /* Get the old TSS selector from the guest's task register */ ot_sel = GETREG(ctx, vcpu, VM_REG_GUEST_TR); if (ISLDT(ot_sel) || IDXSEL(ot_sel) == 0) { /* * This might happen if a task switch was attempted without * ever loading the task register with LTR. In this case the * TR would contain the values from power-on: * (sel = 0, base = 0, limit = 0xffff). */ sel_exception(ctx, vcpu, IDT_TS, ot_sel, task_switch->ext); goto done; } /* Get the old TSS base and limit from the guest's task register */ error = vm_get_desc(ctx, vcpu, VM_REG_GUEST_TR, &ot_base, &ot_lim, &access); assert(error == 0); assert(!SEG_DESC_UNUSABLE(access) && SEG_DESC_PRESENT(access)); ot_type = SEG_DESC_TYPE(access); assert(ot_type == SDT_SYS386BSY || ot_type == SDT_SYS286BSY); /* Fetch the old TSS descriptor */ error = read_tss_descriptor(ctx, vcpu, task_switch, ot_sel, &ot_desc); CHKERR(error); /* Get the old TSS */ error = vm_copy_setup(ctx, vcpu, &sup_paging, ot_base, minlimit + 1, PROT_READ | PROT_WRITE, ot_iov, nitems(ot_iov)); CHKERR(error); vm_copyin(ctx, vcpu, ot_iov, &oldtss, minlimit + 1); /* * Clear the busy bit in the old TSS descriptor if the task switch * due to an IRET or JMP instruction. */ if (reason == TSR_IRET || reason == TSR_JMP) { ot_desc.sd_type &= ~0x2; error = desc_table_write(ctx, vcpu, &sup_paging, ot_sel, &ot_desc); CHKERR(error); } if (nt_type == SDT_SYS286BSY || nt_type == SDT_SYS286TSS) { fprintf(stderr, "Task switch to 16-bit TSS not supported\n"); return (VMEXIT_ABORT); } /* Save processor state in old TSS */ tss32_save(ctx, vcpu, task_switch, eip, &oldtss, ot_iov); /* * If the task switch was triggered for any reason other than IRET * then set the busy bit in the new TSS descriptor. */ if (reason != TSR_IRET) { nt_desc.sd_type |= 0x2; error = desc_table_write(ctx, vcpu, &sup_paging, nt_sel, &nt_desc); CHKERR(error); } /* Update task register to point at the new TSS */ SETREG(ctx, vcpu, VM_REG_GUEST_TR, nt_sel); /* Update the hidden descriptor state of the task register */ nt = usd_to_seg_desc(&nt_desc); update_seg_desc(ctx, vcpu, VM_REG_GUEST_TR, &nt); /* Set CR0.TS */ cr0 = GETREG(ctx, vcpu, VM_REG_GUEST_CR0); SETREG(ctx, vcpu, VM_REG_GUEST_CR0, cr0 | CR0_TS); /* * We are now committed to the task switch. Any exceptions encountered * after this point will be handled in the context of the new task and * the saved instruction pointer will belong to the new task. */ - vmexit->rip = newtss.tss_eip; - assert(vmexit->inst_length == 0); + error = vm_set_register(ctx, vcpu, VM_REG_GUEST_RIP, newtss.tss_eip); + assert(error == 0); /* Load processor state from new TSS */ error = tss32_restore(ctx, vcpu, task_switch, ot_sel, &newtss, nt_iov); CHKERR(error); /* * Section "Interrupt Tasks" in Intel SDM, Vol 3: if an exception * caused an error code to be generated, this error code is copied * to the stack of the new task. */ if (task_switch->errcode_valid) { assert(task_switch->ext); assert(task_switch->reason == TSR_IDT_GATE); error = push_errcode(ctx, vcpu, &task_switch->paging, nt_type, task_switch->errcode); CHKERR(error); } /* * Treatment of virtual-NMI blocking if NMI is delivered through * a task gate. * * Section "Architectural State Before A VM Exit", Intel SDM, Vol3: * If the virtual NMIs VM-execution control is 1, VM entry injects * an NMI, and delivery of the NMI causes a task switch that causes * a VM exit, virtual-NMI blocking is in effect before the VM exit * commences. * * Thus, virtual-NMI blocking is in effect at the time of the task * switch VM exit. */ /* * Treatment of virtual-NMI unblocking on IRET from NMI handler task. * * Section "Changes to Instruction Behavior in VMX Non-Root Operation" * If "virtual NMIs" control is 1 IRET removes any virtual-NMI blocking. * This unblocking of virtual-NMI occurs even if IRET causes a fault. * * Thus, virtual-NMI blocking is cleared at the time of the task switch * VM exit. */ /* * If the task switch was triggered by an event delivered through * the IDT then extinguish the pending event from the vcpu's * exitintinfo. */ if (task_switch->reason == TSR_IDT_GATE) { error = vm_set_intinfo(ctx, vcpu, 0); assert(error == 0); } /* * XXX should inject debug exception if 'T' bit is 1 */ done: return (VMEXIT_CONTINUE); } Index: user/marcel/libvdsk/bhyve/xmsr.c =================================================================== --- user/marcel/libvdsk/bhyve/xmsr.c (revision 279574) +++ user/marcel/libvdsk/bhyve/xmsr.c (revision 279575) @@ -1,224 +1,233 @@ /*- * Copyright (c) 2011 NetApp, 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 NETAPP, INC ``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 NETAPP, INC 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. * * $FreeBSD$ */ #include __FBSDID("$FreeBSD$"); #include #include #include #include #include #include #include #include #include "xmsr.h" static int cpu_vendor_intel, cpu_vendor_amd; int emulate_wrmsr(struct vmctx *ctx, int vcpu, uint32_t num, uint64_t val) { if (num == MSR_PAT) return 0; if (cpu_vendor_intel) { switch (num) { case 0xd04: /* Sandy Bridge uncore PMCs */ case 0xc24: return (0); case MSR_BIOS_UPDT_TRIG: return (0); case MSR_BIOS_SIGN: return (0); default: break; } } else if (cpu_vendor_amd) { switch (num) { case MSR_HWCR: /* * Ignore writes to hardware configuration MSR. */ return (0); case MSR_NB_CFG1: case MSR_IC_CFG: return (0); /* Ignore writes */ case MSR_PERFEVSEL0: case MSR_PERFEVSEL1: case MSR_PERFEVSEL2: case MSR_PERFEVSEL3: /* Ignore writes to the PerfEvtSel MSRs */ return (0); case MSR_K7_PERFCTR0: case MSR_K7_PERFCTR1: case MSR_K7_PERFCTR2: case MSR_K7_PERFCTR3: /* Ignore writes to the PerfCtr MSRs */ return (0); case MSR_P_STATE_CONTROL: /* Ignore write to change the P-state */ return (0); default: break; } } return (-1); } int emulate_rdmsr(struct vmctx *ctx, int vcpu, uint32_t num, uint64_t *val) { int error = 0; if (cpu_vendor_intel) { switch (num) { case MSR_BIOS_SIGN: case MSR_IA32_PLATFORM_ID: case MSR_PKG_ENERGY_STATUS: case MSR_PP0_ENERGY_STATUS: case MSR_PP1_ENERGY_STATUS: case MSR_DRAM_ENERGY_STATUS: *val = 0; break; case MSR_RAPL_POWER_UNIT: /* * Use the default value documented in section * "RAPL Interfaces" in Intel SDM vol3. */ *val = 0x000a1003; break; default: error = -1; break; } } else if (cpu_vendor_amd) { switch (num) { case MSR_BIOS_SIGN: *val = 0; break; case MSR_HWCR: /* * Bios and Kernel Developer's Guides for AMD Families * 12H, 14H, 15H and 16H. */ *val = 0x01000010; /* Reset value */ *val |= 1 << 9; /* MONITOR/MWAIT disable */ break; case MSR_NB_CFG1: case MSR_IC_CFG: /* * The reset value is processor family dependent so * just return 0. */ *val = 0; break; case MSR_PERFEVSEL0: case MSR_PERFEVSEL1: case MSR_PERFEVSEL2: case MSR_PERFEVSEL3: /* * PerfEvtSel MSRs are not properly virtualized so just * return zero. */ *val = 0; break; case MSR_K7_PERFCTR0: case MSR_K7_PERFCTR1: case MSR_K7_PERFCTR2: case MSR_K7_PERFCTR3: /* * PerfCtr MSRs are not properly virtualized so just * return zero. */ *val = 0; break; case MSR_SMM_ADDR: case MSR_SMM_MASK: /* * Return the reset value defined in the AMD Bios and * Kernel Developer's Guide. */ *val = 0; break; case MSR_P_STATE_LIMIT: case MSR_P_STATE_CONTROL: case MSR_P_STATE_STATUS: case MSR_P_STATE_CONFIG(0): /* P0 configuration */ *val = 0; break; + /* + * OpenBSD guests test bit 0 of this MSR to detect if the + * workaround for erratum 721 is already applied. + * http://support.amd.com/TechDocs/41322_10h_Rev_Gd.pdf + */ + case 0xC0011029: + *val = 1; + break; + default: error = -1; break; } } else { error = -1; } return (error); } int init_msr(void) { int error; u_int regs[4]; char cpu_vendor[13]; do_cpuid(0, regs); ((u_int *)&cpu_vendor)[0] = regs[1]; ((u_int *)&cpu_vendor)[1] = regs[3]; ((u_int *)&cpu_vendor)[2] = regs[2]; cpu_vendor[12] = '\0'; error = 0; if (strcmp(cpu_vendor, "AuthenticAMD") == 0) { cpu_vendor_amd = 1; } else if (strcmp(cpu_vendor, "GenuineIntel") == 0) { cpu_vendor_intel = 1; } else { fprintf(stderr, "Unknown cpu vendor \"%s\"\n", cpu_vendor); error = -1; } return (error); } Index: user/marcel/libvdsk/bhyve =================================================================== --- user/marcel/libvdsk/bhyve (revision 279574) +++ user/marcel/libvdsk/bhyve (revision 279575) Property changes on: user/marcel/libvdsk/bhyve ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,2 ## Merged /projects/clang350-import/usr.sbin/bhyve:r274961-276476 Merged /head/usr.sbin/bhyve:r276399-279574 Index: user/marcel/libvdsk/bhyveload =================================================================== --- user/marcel/libvdsk/bhyveload (revision 279574) +++ user/marcel/libvdsk/bhyveload (revision 279575) Property changes on: user/marcel/libvdsk/bhyveload ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,2 ## Merged /projects/clang350-import/usr.sbin/bhyveload:r274961-276476 Merged /head/usr.sbin/bhyveload:r276399-279574