From: Heiko Carstens <[EMAIL PROTECTED]> Add interface which allows a process to start a virtual machine.
To keep things easy each thread group is allowed to have only one virtual machine and each thread of the thread group can only control one virtual cpu of the virtual machine. All the information about the virtual machines/cpus can be found via the thread_info structures of the participating threads. This patch adds three new s390 specific system calls: long sys_s390host_add_cpu(unsigned long addr, unsigned long flags, struct sie_block __user *sie_template) Adds a new cpu to a the virtual machine that belongs to the current thread group. If no virtual machine exists it will be created. In addition two pages will be allocated and mapped at <addr> into the address space of the process. These two pages are used so user space and kernel space can easily exchange/modify the state of the corresponding virtual cpu without a ton of copy_from/to_user calls. The sie_template is a pointer to a data structure that contains initial information how the virtual cpu should be setup. The resulting block will be used as a parameter to issue the sie (start interpretive execution) instruction which starts a virtual cpu. int sys_s390host_remove_cpu(void) Removes a virtual cpu from a virtual machine. int sys_s390host_sie(unsigned long action) Starts / re-enters the virtual cpu of the virtual machine that the thread belongs to, if any. Please note that this patch is nothing more than a proof-of-concept and may contain quite a few bugs. Since we want to convert to use kvm instead, most of this will be dropped anyway. But maybe this is of interest for others as well. Signed-off-by: Heiko Carstens <[EMAIL PROTECTED]> Signed-off-by: Carsten Otte <[EMAIL PROTECTED]> --- arch/s390/Kconfig | 7 arch/s390/Makefile | 2 arch/s390/host/Makefile | 5 arch/s390/host/s390_intercept.c | 42 ++++ arch/s390/host/s390host.c | 418 ++++++++++++++++++++++++++++++++++++++++ arch/s390/host/s390host.h | 16 + arch/s390/host/sie64a.S | 38 +++ arch/s390/kernel/asm-offsets.c | 2 arch/s390/kernel/process.c | 15 + arch/s390/kernel/setup.c | 4 arch/s390/kernel/syscalls.S | 3 include/asm-s390/sie64.h | 279 ++++++++++++++++++++++++++ include/asm-s390/thread_info.h | 8 include/asm-s390/unistd.h | 5 kernel/sys_ni.c | 3 15 files changed, 842 insertions(+), 5 deletions(-) Index: linux-2.6.21/arch/s390/kernel/asm-offsets.c =================================================================== --- linux-2.6.21.orig/arch/s390/kernel/asm-offsets.c +++ linux-2.6.21/arch/s390/kernel/asm-offsets.c @@ -44,5 +44,7 @@ int main(void) DEFINE(__SF_BACKCHAIN, offsetof(struct stack_frame, back_chain),); DEFINE(__SF_GPRS, offsetof(struct stack_frame, gprs),); DEFINE(__SF_EMPTY, offsetof(struct stack_frame, empty1),); + BLANK(); + DEFINE(__SIE_USER_gprs, offsetof(struct sie_user, gprs),); return 0; } Index: linux-2.6.21/arch/s390/kernel/syscalls.S =================================================================== --- linux-2.6.21.orig/arch/s390/kernel/syscalls.S +++ linux-2.6.21/arch/s390/kernel/syscalls.S @@ -322,3 +322,6 @@ NI_SYSCALL /* 310 sys_move_pages * SYSCALL(sys_getcpu,sys_getcpu,sys_getcpu_wrapper) SYSCALL(sys_epoll_pwait,sys_epoll_pwait,compat_sys_epoll_pwait_wrapper) SYSCALL(sys_utimes,sys_utimes,compat_sys_utimes_wrapper) +SYSCALL(sys_ni_syscall,sys_s390host_add_cpu,sys_ni_syscall) +SYSCALL(sys_ni_syscall,sys_s390host_remove_cpu,sys_ni_syscall) +SYSCALL(sys_ni_syscall,sys_s390host_sie,sys_ni_syscall) Index: linux-2.6.21/arch/s390/host/Makefile =================================================================== --- /dev/null +++ linux-2.6.21/arch/s390/host/Makefile @@ -0,0 +1,5 @@ +# +# Makefile for the s390host components. +# + +obj-$(CONFIG_S390_HOST) += s390host.o sie64a.o s390_intercept.o Index: linux-2.6.21/arch/s390/host/sie64a.S =================================================================== --- /dev/null +++ linux-2.6.21/arch/s390/host/sie64a.S @@ -0,0 +1,38 @@ +/* + * arch/s390/host/sie64a.S + * low level sie call + * + * Copyright IBM Corp. 2007 + * Author(s): Heiko Carstens <[EMAIL PROTECTED]> + * License : GPL + */ + +#include <linux/errno.h> +#include <asm/asm-offsets.h> + +SP_R6 = 6 * 8 # offset into stackframe + + .globl sie64a +sie64a: + stmg %r6,%r15,SP_R6(%r15) # save register on entry + lgr %r14,%r2 # pointer to program parms + aghi %r2,4096 + lmg %r0,%r13,__SIE_USER_gprs(%r2) # load guest gprs 0-13 +sie_inst: + sie 0(%r14) + aghi %r14,4096 + stmg %r0,%r13,__SIE_USER_gprs(%r14) # save guest gprs 0-13 + lghi %r2,0 + lmg %r6,%r15,SP_R6(%r15) + br %r14 + +sie_err: + aghi %r14,4096 + stmg %r0,%r13,__SIE_USER_gprs(%r14) # save guest gprs 0-13 + lghi %r2,-EFAULT + lmg %r6,%r15,SP_R6(%r15) + br %r14 + + .section __ex_table,"a" + .quad sie_inst,sie_err + .previous Index: linux-2.6.21/arch/s390/host/s390host.c =================================================================== --- /dev/null +++ linux-2.6.21/arch/s390/host/s390host.c @@ -0,0 +1,418 @@ +/* + * s390host.c -- hosting zSeries Linux virtual engines + * + * Copyright IBM Corp. 2007 + * Author(s): Carsten Otte <[EMAIL PROTECTED]>, + * Christian Borntraeger <[EMAIL PROTECTED]>, + * Heiko Carstens <[EMAIL PROTECTED]> + * License : GPL + */ + +#include <linux/pagemap.h> +#include <linux/module.h> +#include <linux/fs.h> +#include <linux/mm.h> +#include <linux/init.h> +#include <linux/file.h> +#include <linux/mman.h> +#include <linux/mutex.h> +#include <asm/uaccess.h> +#include <asm/processor.h> +#include <asm/tlbflush.h> +#include <asm/semaphore.h> +#include <asm/sie64.h> +#include "s390host.h" + +static int s390host_do_action(unsigned long, struct sie_io *); + +static DEFINE_MUTEX(s390host_init_mutex); + +static void s390host_get_data(struct s390host_data *data) +{ + atomic_inc(&data->count); +} + +void s390host_put_data(struct s390host_data *data) +{ + int cpu; + + if (atomic_dec_return(&data->count)) + return; + + for (cpu = 0; cpu < S390HOST_MAX_CPUS; cpu++) + if (data->sie_io[cpu]) + free_page((unsigned long)data->sie_io[cpu]); + + if (data->sca_block) + free_page((unsigned long)data->sca_block); + + kfree(data); +} + +static void s390host_vma_close(struct vm_area_struct *vma) +{ + s390host_put_data(vma->vm_private_data); +} + +static struct page *s390host_vma_nopage(struct vm_area_struct *vma, + unsigned long address, int *type) +{ + return NOPAGE_SIGBUS; +} + +static struct vm_operations_struct s390host_vmops = { + .close = s390host_vma_close, + .nopage = s390host_vma_nopage, +}; + +static struct s390host_data *get_s390host_context(void) +{ + struct thread_info *tif; + struct sca_block *sca_block = NULL; + struct s390host_data *data = NULL; + struct task_struct *tsk; + + /* zlh context for current thread already created? */ + tif = current_thread_info(); + if (tif->s390host_data) + return tif->s390host_data; + + /* zlh context in thread group available? */ + write_lock_irq(&tasklist_lock); + tsk = next_thread(current); + for (; tsk != current; tsk = next_thread(tsk)) { + data = tsk->thread_info->s390host_data; + if (data) { + s390host_get_data(data); + tif->s390host_data = data; + break; + } + } + write_unlock_irq(&tasklist_lock); + + if (data) + return data; + + /* create new context */ + data = kzalloc(sizeof(*data), GFP_KERNEL); + + if (!data) + return NULL; + + sca_block = (struct sca_block *)get_zeroed_page(GFP_KERNEL); + + if (!sca_block) { + kfree(data); + return NULL; + } + + data->sca_block = sca_block; + tif->s390host_data = data; + s390host_get_data(data); + + return data; +} + +static unsigned long +s390host_create_io_area(unsigned long addr, unsigned long flags, + unsigned long io_addr, struct s390host_data *data) +{ + struct mm_struct *mm = current->mm; + struct vm_area_struct *vma; + unsigned long ret; + + flags &= MAP_FIXED; + addr = get_unmapped_area(NULL, addr, 2 * PAGE_SIZE, 0, flags); + + if (addr & ~PAGE_MASK) + return addr; + + vma = kmem_cache_zalloc(vm_area_cachep, GFP_KERNEL); + + if (!vma) + return -ENOMEM; + + vma->vm_mm = mm; + vma->vm_start = addr; + vma->vm_end = addr + 2 * PAGE_SIZE; + vma->vm_flags = VM_READ | VM_MAYREAD | VM_IO | VM_RESERVED; + vma->vm_flags |= VM_SHARED | VM_MAYSHARE | VM_DONTCOPY; + +#if 1 /* FIXME: write access until sys_s390host_sie interface is final */ + vma->vm_flags |= VM_WRITE | VM_MAYWRITE; +#endif + + vma->vm_page_prot = protection_map[vma->vm_flags & 0xf]; + vma->vm_private_data = data; + vma->vm_ops = &s390host_vmops; + + down_write(&mm->mmap_sem); + ret = insert_vm_struct(mm, vma); + if (ret) { + kmem_cache_free(vm_area_cachep, vma); + goto out; + } + s390host_get_data(data); + mm->total_vm += 2; + vm_insert_page(vma, addr, virt_to_page(io_addr)); + + ret = split_vma(mm, vma, addr + PAGE_SIZE, 0); + if (ret) + goto out; + s390host_get_data(data); + + vma = find_vma(mm, addr + PAGE_SIZE); + vma->vm_flags |= VM_WRITE | VM_MAYWRITE; + vma->vm_page_prot = protection_map[vma->vm_flags & 0xf]; + vm_insert_page(vma, addr + PAGE_SIZE, + virt_to_page(io_addr + PAGE_SIZE)); + ret = addr; +out: + up_write(&mm->mmap_sem); + return ret; +} + +long sys_s390host_add_cpu(unsigned long addr, unsigned long flags, + struct sie_block __user *sie_template) +{ + struct sie_block *sie_block; + struct sie_io *sie_io; + struct sca_block *sca_block; + struct s390host_data *data = NULL; + unsigned long ret; + __u16 cpu; + + if (current_thread_info()->sie_cpu != -1) + return -EINVAL; + + if (copy_from_user(&cpu, &sie_template->icpua, sizeof(u16))) + return -EFAULT; + + if (cpu >= S390HOST_MAX_CPUS) + return -EINVAL; + + mutex_lock(&s390host_init_mutex); + + data = get_s390host_context(); + if (!data) { + ret = -ENOMEM; + goto out_err; + } + + sca_block = data->sca_block; + if (sca_block->mcn & (1UL << (S390HOST_MAX_CPUS - 1 - cpu))) { + ret = -EINVAL; + goto out_err; + } + + if (!data->sie_io[cpu]) { + unsigned long tmp; + + /* allocate two pages: 1st is r/o 2nd r/w area */ + tmp = __get_free_pages(GFP_KERNEL, 1); + if (!tmp) { + ret = -ENOMEM; + goto out_err; + } + split_page(virt_to_page(tmp), 1); + data->sie_io[cpu] = (struct sie_io *)tmp; + } + + sie_io = data->sie_io[cpu]; + memset(sie_io, 0, 2 * PAGE_SIZE); + + sie_block = &sie_io->sie_kernel.sie_block; + sca_block->cpu[cpu].sda = (__u64)sie_block; + + if (copy_from_user(sie_block, sie_template, sizeof(struct sie_block))) { + ret = -EFAULT; + goto out_err; + } + sie_block->icpua = cpu; + + ret = s390host_create_io_area(addr, flags, (unsigned long)sie_io, data); + + if (ret & ~PAGE_MASK) + goto out_err; + + sca_block->mcn |= 1UL << (S390HOST_MAX_CPUS - 1 - cpu); + sie_block->scaoh = (__u32)(((__u64)sca_block) >> 32); + sie_block->scaol = (__u32)(__u64)sca_block; + current_thread_info()->sie_cpu = cpu; + goto out; +out_err: + if (data) + s390host_put_data(data); +out: + mutex_unlock(&s390host_init_mutex); + return ret; +} + +int sys_s390host_remove_cpu(void) +{ + struct sca_block *sca_block; + int cpu; + + cpu = current_thread_info()->sie_cpu; + if (cpu == -1) + return -EINVAL; + + mutex_lock(&s390host_init_mutex); + sca_block = current_thread_info()->s390host_data->sca_block; + sca_block->mcn &= ~(1UL << (S390HOST_MAX_CPUS - 1 - cpu)); + current_thread_info()->sie_cpu = -1; + mutex_unlock(&s390host_init_mutex); + return 0; +} + +int sys_s390host_sie(unsigned long action) +{ + struct sie_kernel *sie_kernel; + struct sie_user *sie_user; + struct sie_io *sie_io; + int cpu; + int ret = 0; + + cpu = current_thread_info()->sie_cpu; + if (cpu == -1) + return -EINVAL; + + sie_io = current_thread_info()->s390host_data->sie_io[cpu]; + + if (action) + ret = s390host_do_action(action, sie_io); + if (ret) + goto out_err; + sie_kernel = &sie_io->sie_kernel; + sie_user = &sie_io->sie_user; + + save_fp_regs(&sie_kernel->host_fpregs); + save_access_regs(sie_kernel->host_acrs); + sie_user->guest_fpregs.fpc &= FPC_VALID_MASK; + restore_fp_regs(&sie_user->guest_fpregs); + restore_access_regs(sie_user->guest_acrs); + memcpy(&sie_kernel->sie_block.gg14, &sie_user->gprs[14], 16); +again: + if (need_resched()) + schedule(); + + sie_kernel->sie_block.icptcode = 0; + ret = sie64a(sie_kernel); + if (ret) + goto out; + + if (signal_pending(current)) { + ret = -EINTR; + goto out; + } + + ret = s390host_handle_intercept(sie_kernel); + + /* intercept reason was handled, enter SIE again */ + if (!ret) + goto again; + + /* if kernel cannot hanle intercept, pass to the user */ + if (ret == -ENOTSUPP) + ret = 0; + +out: + memcpy(&sie_user->gprs[14], &sie_kernel->sie_block.gg14, 16); + save_fp_regs(&sie_user->guest_fpregs); + save_access_regs(sie_user->guest_acrs); + restore_fp_regs(&sie_kernel->host_fpregs); + restore_access_regs(sie_kernel->host_acrs); +out_err: + return ret; +} + +static void s390host_vsmxm_local_update(struct sie_io *sie_io) +{ + struct sie_kernel *local_sie_kernel; + struct sie_user *sie_user; + atomic_t *cpuflags; + int old, new; + + mutex_lock(&s390host_init_mutex); + + sie_user = &sie_io->sie_user; + local_sie_kernel = &sie_io->sie_kernel; + + cpuflags = &local_sie_kernel->sie_block.cpuflags; + do { + old = atomic_read(cpuflags); + new = old | sie_user->vsmxm_or_local; + new &= sie_user->vsmxm_and_local; + } while (atomic_cmpxchg(cpuflags, old, new) != old); + + mutex_unlock(&s390host_init_mutex); + return; +} + +static int s390host_vsmxm_dist_update(struct sie_io *sie_io) +{ + struct sie_kernel *dist_sie_kernel; + struct sie_user *sie_user; + struct sca_block *sca_block; + struct thread_info *tif; + atomic_t *cpuflags; + int cpu; + int old, new; + int rc = -EINVAL; + + mutex_lock(&s390host_init_mutex); + + sie_user = &sie_io->sie_user; + cpu = sie_user->vsmxm_cpuid; + + if (cpu >= S390HOST_MAX_CPUS) + goto out; + + tif = current_thread_info(); + sca_block = tif->s390host_data->sca_block; + if (!(sca_block->mcn & (1UL << (S390HOST_MAX_CPUS - 1 - cpu)))) + goto out; + + dist_sie_kernel = &((tif->s390host_data->sie_io[cpu])->sie_kernel); + + cpuflags = &dist_sie_kernel->sie_block.cpuflags; + do { + old = atomic_read(cpuflags); + new = old | sie_user->vsmxm_or; + new &= sie_user->vsmxm_and; + } while (atomic_cmpxchg(cpuflags, old, new) != old); + + rc = 0; +out: + mutex_unlock(&s390host_init_mutex); + return rc; +} + +static int s390host_do_action(unsigned long action, struct sie_io *sie_io) +{ + void *src; + void *dest; + int rc = 0; + + if (action & SIE_BLOCK_UPDATE) { + src = &(sie_io->sie_user.sie_block); + dest = &(sie_io->sie_kernel.sie_block); + + memcpy(dest + 4, src + 4, 88); + memcpy(dest + 96, src + 96, 4); + memcpy(dest + 104, src + 104, 408); + } + + if (action & SIE_UPDATE_PSW) + sie_io->sie_kernel.sie_block.psw.gpsw = sie_io->sie_user.psw; + + if (action & SIE_FLUSH_TLB) + flush_tlb_mm(current->mm); + + if (action & SIE_VSMXM_LOCAL_UPDATE) + s390host_vsmxm_local_update(sie_io); + + if (action & SIE_VSMXM_DIST_UPDATE) + rc = s390host_vsmxm_dist_update(sie_io); + return rc; +} Index: linux-2.6.21/include/asm-s390/unistd.h =================================================================== --- linux-2.6.21.orig/include/asm-s390/unistd.h +++ linux-2.6.21/include/asm-s390/unistd.h @@ -251,8 +251,11 @@ #define __NR_getcpu 311 #define __NR_epoll_pwait 312 #define __NR_utimes 313 +#define __NR_s390host_add_cpu 314 +#define __NR_s390host_remove_cpu 315 +#define __NR_s390host_sie 316 -#define NR_syscalls 314 +#define NR_syscalls 317 /* * There are some system calls that are not present on 64 bit, some Index: linux-2.6.21/include/asm-s390/sie64.h =================================================================== --- /dev/null +++ linux-2.6.21/include/asm-s390/sie64.h @@ -0,0 +1,279 @@ +/* + * include/asm-s390/sie64.h + * + * Copyright IBM Corp. 2007 + * Author(s): Carsten Otte <[EMAIL PROTECTED]>, + * Christian Borntraeger <[EMAIL PROTECTED]>, + * Heiko Carstens <[EMAIL PROTECTED]> + * + */ + +#ifndef _ASM_S390_SIE64_H +#define _ASM_S390_SIE64_H + +#include <asm/atomic.h> +#include <asm/ptrace.h> //FIXME: psw_t definition needs relocation + +struct sie_block { + atomic_t cpuflags; /* 0x0000 */ + __u32 prefix; /* 0x0004 */ + __u32 :32; /* 0x0008 */ + __u32 :32; /* 0x000c */ + __u64 :64; /* 0x0010 */ + __u64 :64; /* 0x0018 */ + __u64 :64; /* 0x0020 */ + __u64 cputm; /* 0x0028 */ + __u64 ckc; /* 0x0030 */ + __u64 epoch; /* 0x0038 */ + __u8 svcnn :1, /* 0x0040 */ + svc1c :1, + svc2c :1, + svc3c :1, + :4; + __u8 svc1n; /* 0x0041 */ + __u8 svc2n; /* 0x0042 */ + __u8 svc3n; /* 0x0043 */ + __u16 lctl0 :1, /* 0x0044 */ + lctl1 :1, + lctl2 :1, + lctl3 :1, + lctl4 :1, + lctl5 :1, + lctl6 :1, + lctl7 :1, + lctl8 :1, + lctl9 :1, + lctla :1, + lctlb :1, + lctlc :1, + lctld :1, + lctle :1, + lctlf :1; + __s16 icpua; /* 0x0046 */ + __u32 icpop :1, /* 0x0048 */ + icpro :1, + icprg :1, + :4, + icipte :1, + :1, /* 0x0049 */ + iclpsw :1, + icptlb :1, + icssm :1, + icbsa :1, + icstctl :1, + icstnsm :1, + icstosm :1, + icstck :1, /* 0x004a */ + iciske :1, + icsske :1, + icrrbe :1, + icpc :1, + icpt :1, + ictprot :1, + iclasp :1, + :1, /* 0x004b */ + icstpt :1, + icsckc :1, + :1, + icpr :1, + icbakr :1, + icpg :1, + :1; + __u32 ecext :1, /* 0x004c */ + ecint :1, + ecwait :1, + ecsigp :1, + ecalt :1, + ecio2 :1, + :1, + ecmvp :1; + __u8 eca1; /* 0x004d */ + __u8 eca2; /* 0x004e */ + __u8 eca3; /* 0x004f */ + __u8 icptcode; /* 0x0050 */ + __u8 :6, /* 0x0051 */ + icif :1, + icex :1; + __u16 ihcpu; /* 0x0052 */ + __u16 :16; /* 0x0054 */ + struct { + union { + __u16 ipa; /* 0x0056 */ + __u16 inst; /* 0x0056 */ + struct { + union { + __u8 ipa0; /* 0x0056 */ + __u8 viwho; /* 0x0056 */ + }; + union { + __u8 ipa1; /* 0x0057 */ + __u8 viwhen; /* 0x0057 */ + }; + }; + }; + union { + __u32 ipb; /* 0x0058 */ + struct { + union { + __u16 ipbh0; /* 0x0058 */ + __u16 viwhy; /* 0x0058 */ + struct { + __u8 ipb0; /* 0x0058 */ + __u8 ipb1; /* 0x0059 */ + }; + }; + union { + __u16 ipbh1; /* 0x005a */ + struct { + __u8 ipb2; /* 0x005a */ + __u8 ipb3; /* 0x005b */ + }; + }; + }; + }; + } __attribute__((packed)); + __u32 scaoh; /* 0x005c */ + union { + __u32 rcp; /* 0x0060 */ + struct { + __u8 ska :1, /* 0x0060 */ + skaip :1, + :6; + __u8 ecb :8; /* 0x0061 */ + __u8 :3, /* 0x0062 */ + cpby :1, + :4; + __u8 :8; /* 0x0063 */ + }; + }; + __u32 scaol; /* 0x0064 */ + __u32 :32; /* 0x0068 */ + union { + __u32 todpr; /* 0x006c */ + struct { + __u16 :16; /* 0x006c */ + __u16 todpf; /* 0x006e */ + }; + } __attribute__((packed)); + __u32 gisa; /* 0x0070 */ + __u32 iopct; /* 0x0074 */ + __u32 rsvd3; /* 0x0078 */ + __u32 :32; /* 0x007c */ + __u64 gmsor; /* 0x0080 */ + __u64 gmslm; /* 0x0088 */ + union { + psw_t gpsw; /* 0x0090 */ + struct { + __u64 pswh; /* 0x0090 */ + + __u64 pswl; /* 0x0098 */ + }; + } psw; + __u64 gg14; /* 0x00a0 */ + __u64 gg15; /* 0x00a8 */ + __u64 :64; /* 0x00b0 */ + __u64 :16, /* 0x00b8 */ + xso :24, + xsl :24; + union { + __u8 uzp0[56]; /* 0x00c0 */ + struct { + __u32 exmsf; /* 0x00c0 */ + union { + __u32 iexcf; /* 0x00c4 */ + struct { + __u16 iexca; /* 0x00c4 */ + __u16 iexcd; /* 0x00c6 */ + }; + }; + __u16 svcil; /* 0x00c8 */ + __u16 svcnt; /* 0x00ca */ + __u16 iprcl; /* 0x00cc */ + __u16 iprcc; /* 0x00ce */ + __u32 itrad; /* 0x00d0 */ + __u32 imncl; /* 0x00d4 */ + __u64 gpera; /* 0x00d8 */ + __u8 excpar; /* 0x00e0 */ + __u8 perar; /* 0x00e1 */ + __u8 oprid; /* 0x00e2 */ + __u8 :8; /* 0x00e3 */ + __u32 :32; /* 0x00e4 */ + __u64 gtrad; /* 0x00e8 */ + __u32 :32; /* 0x00f0 */ + __u32 :32; /* 0x00f4 */ + }; + }; + __u16 :16; /* 0x00f8 */ + __u16 ief; /* 0x00fa */ + __u32 apcbk; /* 0x00fc */ + __u64 gcr[16]; /* 0x0100 */ + __u8 reserved[128]; /* 0x0180 */ +} __attribute__((packed)); + +struct sie_kernel { + struct sie_block sie_block; + s390_fp_regs host_fpregs; + int host_acrs[NUM_ACRS]; +} __attribute__((packed,aligned(4096))); + +#define SIE_UPDATE_PSW (1UL << 0) +#define SIE_FLUSH_TLB (1UL << 1) +#define SIE_ISKE (1UL << 2) +#define SIE_SSKE (1UL << 3) +#define SIE_BLOCK_UPDATE (1UL << 4) +#define SIE_VSMXM_LOCAL_UPDATE (1UL << 5) +#define SIE_VSMXM_DIST_UPDATE (1UL << 6) + +struct sie_skey_parm { + unsigned long sk_reg; + unsigned long sk_addr; +}; + +struct sie_user { + struct sie_block sie_block; + psw_t psw; + unsigned long gprs[NUM_GPRS]; + s390_fp_regs guest_fpregs; + int guest_acrs[NUM_ACRS]; + struct sie_skey_parm iske_parm; + struct sie_skey_parm sske_parm; + int vsmxm_or_local; + int vsmxm_and_local; + int vsmxm_or; + int vsmxm_and; + int vsmxm_cpuid; +} __attribute__((packed,aligned(4096))); + +struct sie_io { + struct sie_kernel sie_kernel; + struct sie_user sie_user; +}; + +struct sca_entry { + atomic_t scn; + __u64 reserved; + __u64 sda; + __u64 reserved2[2]; +}__attribute__((packed)); + +struct sca_block { + __u64 ipte_control; + __u64 reserved[5]; + __u64 mcn; + __u64 reserved2; + struct sca_entry cpu[64]; +}__attribute__((packed)); + +#define S390HOST_MAX_CPUS 64 + +struct s390host_data { + atomic_t count; + struct sie_io *sie_io[S390HOST_MAX_CPUS]; + struct sca_block *sca_block; +}; + +/* function definitions */ +extern int sie64a(struct sie_kernel *); +extern void s390host_put_data(struct s390host_data *); + +#endif /* _ASM_S390_SIE64_H */ Index: linux-2.6.21/arch/s390/Makefile =================================================================== --- linux-2.6.21.orig/arch/s390/Makefile +++ linux-2.6.21/arch/s390/Makefile @@ -85,7 +85,7 @@ LDFLAGS_vmlinux := -e start head-y := arch/s390/kernel/head.o arch/s390/kernel/init_task.o core-y += arch/s390/mm/ arch/s390/kernel/ arch/s390/crypto/ \ - arch/s390/appldata/ arch/s390/hypfs/ + arch/s390/appldata/ arch/s390/hypfs/ arch/s390/host/ libs-y += arch/s390/lib/ drivers-y += drivers/s390/ drivers-$(CONFIG_MATHEMU) += arch/s390/math-emu/ Index: linux-2.6.21/kernel/sys_ni.c =================================================================== --- linux-2.6.21.orig/kernel/sys_ni.c +++ linux-2.6.21/kernel/sys_ni.c @@ -122,6 +122,9 @@ cond_syscall(sys32_sysctl); cond_syscall(ppc_rtas); cond_syscall(sys_spu_run); cond_syscall(sys_spu_create); +cond_syscall(sys_s390host_add_cpu); +cond_syscall(sys_s390host_remove_cpu); +cond_syscall(sys_s390host_sie); /* mmu depending weak syscall entries */ cond_syscall(sys_mprotect); Index: linux-2.6.21/arch/s390/kernel/process.c =================================================================== --- linux-2.6.21.orig/arch/s390/kernel/process.c +++ linux-2.6.21/arch/s390/kernel/process.c @@ -274,12 +274,23 @@ int copy_thread(int nr, unsigned long cl #endif /* CONFIG_64BIT */ /* start new process with ar4 pointing to the correct address space */ p->thread.mm_segment = get_fs(); - /* Don't copy debug registers */ - memset(&p->thread.per_info,0,sizeof(p->thread.per_info)); + /* Don't copy debug registers */ + memset(&p->thread.per_info,0,sizeof(p->thread.per_info)); + p->thread_info->s390host_data = NULL; + p->thread_info->sie_cpu = -1; return 0; } +void free_thread_info(struct thread_info *ti) +{ +#ifdef CONFIG_S390_HOST + if (ti->s390host_data) + s390host_put_data(ti->s390host_data); +#endif + free_pages((unsigned long) (ti),THREAD_ORDER); +} + asmlinkage long sys_fork(struct pt_regs regs) { return do_fork(SIGCHLD, regs.gprs[15], ®s, 0, NULL, NULL); Index: linux-2.6.21/include/asm-s390/thread_info.h =================================================================== --- linux-2.6.21.orig/include/asm-s390/thread_info.h +++ linux-2.6.21/include/asm-s390/thread_info.h @@ -38,6 +38,7 @@ #ifndef __ASSEMBLY__ #include <asm/processor.h> #include <asm/lowcore.h> +#include <asm/sie64.h> /* * low level task data that entry.S needs immediate access to @@ -52,6 +53,8 @@ struct thread_info { unsigned int cpu; /* current CPU */ int preempt_count; /* 0 => preemptable, <0 => BUG */ struct restart_block restart_block; + struct s390host_data *s390host_data; /* s390host data */ + int sie_cpu; /* sie cpu number */ }; /* @@ -67,6 +70,8 @@ struct thread_info { .restart_block = { \ .fn = do_no_restart_syscall, \ }, \ + .s390host_data = NULL, \ + .sie_cpu = 0, \ } #define init_thread_info (init_thread_union.thread_info) @@ -81,7 +86,8 @@ static inline struct thread_info *curren /* thread information allocation */ #define alloc_thread_info(tsk) ((struct thread_info *) \ __get_free_pages(GFP_KERNEL,THREAD_ORDER)) -#define free_thread_info(ti) free_pages((unsigned long) (ti),THREAD_ORDER) + +extern void free_thread_info(struct thread_info *); #endif Index: linux-2.6.21/arch/s390/Kconfig =================================================================== --- linux-2.6.21.orig/arch/s390/Kconfig +++ linux-2.6.21/arch/s390/Kconfig @@ -153,6 +153,7 @@ config S390_SWITCH_AMODE config S390_EXEC_PROTECT bool "Data execute protection" + depends on !S390_HOST select S390_SWITCH_AMODE help This option allows to enable a buffer overflow protection for user @@ -514,6 +515,12 @@ config KEXEC current kernel, and to start another kernel. It is like a reboot but is independent of hardware/microcode support. +config S390_HOST + bool "s390 host support (EXPERIMENTAL)" + depends on 64BIT && EXPERIMENTAL + select S390_SWITCH_AMODE + help + Select this option if you want to host guest Linux images endmenu source "net/Kconfig" Index: linux-2.6.21/arch/s390/kernel/setup.c =================================================================== --- linux-2.6.21.orig/arch/s390/kernel/setup.c +++ linux-2.6.21/arch/s390/kernel/setup.c @@ -394,7 +394,11 @@ static int __init early_parse_ipldelay(c early_param("ipldelay", early_parse_ipldelay); #ifdef CONFIG_S390_SWITCH_AMODE +#ifdef CONFIG_S390_HOST +unsigned int switch_amode = 1; +#else unsigned int switch_amode = 0; +#endif EXPORT_SYMBOL_GPL(switch_amode); static void set_amode_and_uaccess(unsigned long user_amode, Index: linux-2.6.21/arch/s390/host/s390_intercept.c =================================================================== --- /dev/null +++ linux-2.6.21/arch/s390/host/s390_intercept.c @@ -0,0 +1,42 @@ +/* + * s390_intercept.c -- handle SIE intercept codes + * + * Copyright IBM Corp. 2007 + * Author(s): Carsten Otte <[EMAIL PROTECTED]> + */ + +#include <linux/kernel.h> +#include <linux/errno.h> +#include <asm/sie64.h> +#include <linux/pagemap.h> +#include "s390host.h" + +static int s390host_handle_validity (struct sie_kernel *sie_kernel) +{ + if (sie_kernel->sie_block.viwhy == 0x37) { + //debug message here + fault_in_pages_writeable((void*)0 + S390HOST_ORIGIN, + PAGE_SIZE); + fault_in_pages_writeable((void*)(unsigned long) + sie_kernel->sie_block.prefix+ + S390HOST_ORIGIN, 2*PAGE_SIZE); + return 0; + } + // debug message here + return -ENOTSUPP; +} + +int s390host_handle_intercept(struct sie_kernel *sie_kernel) +{ + switch (sie_kernel->sie_block.icptcode) { + case 0x00: + case 0x24: + return 0; + case 0x20: + return s390host_handle_validity(sie_kernel); + default: + // debug message here + return -ENOTSUPP; + } +} + Index: linux-2.6.21/arch/s390/host/s390host.h =================================================================== --- /dev/null +++ linux-2.6.21/arch/s390/host/s390host.h @@ -0,0 +1,16 @@ +/* + * s390host.h -- hosting zSeries Linux virtual engines + * + * Copyright IBM Corp. 2007 + * Author(s): Carsten Otte <[EMAIL PROTECTED]>, + * Christian Borntraeger <[EMAIL PROTECTED]>, + * Heiko Carstens <[EMAIL PROTECTED]> + */ + +#ifndef __S390HOST_H +#define __S390HOST_H +#include <asm/sie64.h> +#define S390HOST_ORIGIN 0 + +int s390host_handle_intercept(struct sie_kernel *sie_kernel); +#endif // defined __S390HOST_H ------------------------------------------------------------------------- This SF.net email is sponsored by DB2 Express Download DB2 Express C - the FREE version of DB2 express and take control of your XML. No limits. Just data. Click to get it now. http://sourceforge.net/powerbar/db2/ _______________________________________________ kvm-devel mailing list kvm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/kvm-devel