void smp_sched_handler(void)
{
unsigned flgs;
unsigned cpu = cpuid;
flgs = sched_ipi_data[cpu].flags;
if (flgs) {
struct proc * p;
p = (struct proc *)sched_ipi_data[cpu].data;
if (flgs & SCHED_IPI_STOP_PROC) {
RTS_SET(p, RTS_PROC_STOP);
}
if (flgs & SCHED_IPI_SAVE_CTX) {
/* all context has been saved already, FPU remains */
if (proc_used_fpu(p) &&
get_cpulocal_var(fpu_owner) == p) {
disable_fpu_exception();
save_local_fpu(p, FALSE /*retain*/);
/* we're preparing to migrate somewhere else */
release_fpu(p);
}
}
if (flgs & SCHED_IPI_VM_INHIBIT) {
RTS_SET(p, RTS_VMINHIBIT);
}
}
__insn_barrier();
sched_ipi_data[cpu].flags = 0;
}
void smp_schedule_vminhibit(struct proc * p)
{
if (proc_is_runnable(p))
smp_schedule_sync(p, SCHED_IPI_VM_INHIBIT);
else
RTS_SET(p, RTS_VMINHIBIT);
assert(RTS_ISSET(p, RTS_VMINHIBIT));
}
void smp_schedule_stop_proc(struct proc * p)
{
if (proc_is_runnable(p))
smp_schedule_sync(p, SCHED_IPI_STOP_PROC);
else
RTS_SET(p, RTS_PROC_STOP);
assert(RTS_ISSET(p, RTS_PROC_STOP));
}
/*===========================================================================*
* do_runctl *
*===========================================================================*/
int do_runctl(struct proc * caller, message * m_ptr)
{
/* Control a process's RTS_PROC_STOP flag. Used for process management.
* If the process is queued sending a message or stopped for system call
* tracing, and the RC_DELAY request flag is given, set MF_SIG_DELAY instead
* of RTS_PROC_STOP, and send a SIGSNDELAY signal later when the process is done
* sending (ending the delay). Used by PM for safe signal delivery.
*/
int proc_nr, action, flags;
register struct proc *rp;
/* Extract the message parameters and do sanity checking. */
if (!isokendpt(m_ptr->RC_ENDPT, &proc_nr)) return(EINVAL);
if (iskerneln(proc_nr)) return(EPERM);
rp = proc_addr(proc_nr);
action = m_ptr->RC_ACTION;
flags = m_ptr->RC_FLAGS;
/* Is the target sending or syscall-traced? Then set MF_SIG_DELAY instead.
* Do this only when the RC_DELAY flag is set in the request flags field.
* The process will not become runnable before PM has called SYS_ENDKSIG.
* Note that asynchronous messages are not covered: a process using SENDA
* should not also install signal handlers *and* expect POSIX compliance.
*/
if (action == RC_STOP && (flags & RC_DELAY)) {
if (RTS_ISSET(rp, RTS_SENDING) || (rp->p_misc_flags & MF_SC_DEFER))
rp->p_misc_flags |= MF_SIG_DELAY;
if (rp->p_misc_flags & MF_SIG_DELAY)
return (EBUSY);
}
/* Either set or clear the stop flag. */
switch (action) {
case RC_STOP:
#if CONFIG_SMP
/* check if we must stop a process on a different CPU */
if (rp->p_cpu != cpuid) {
smp_schedule_stop_proc(rp);
break;
}
#endif
RTS_SET(rp, RTS_PROC_STOP);
break;
case RC_RESUME:
assert(RTS_ISSET(rp, RTS_PROC_STOP));
RTS_UNSET(rp, RTS_PROC_STOP);
break;
default:
return(EINVAL);
}
return(OK);
}
/*
* This function gets always called only after smp_sched_handler() has been
* already called. It only serves the purpose of acknowledging the IPI and
* preempting the current process if the CPU was not idle.
*/
void smp_ipi_sched_handler(void)
{
struct proc * curr;
ipi_ack();
curr = get_cpulocal_var(proc_ptr);
if (curr->p_endpoint != IDLE) {
RTS_SET(curr, RTS_PREEMPTED);
}
}
/*===========================================================================*
* do_vmctl *
*===========================================================================*/
int do_vmctl(struct proc * caller, message * m_ptr)
{
int proc_nr;
endpoint_t ep = m_ptr->SVMCTL_WHO;
struct proc *p, *rp, **rpp, *target;
if(ep == SELF) { ep = caller->p_endpoint; }
if(!isokendpt(ep, &proc_nr)) {
printf("do_vmctl: unexpected endpoint %d from VM\n", ep);
return EINVAL;
}
p = proc_addr(proc_nr);
switch(m_ptr->SVMCTL_PARAM) {
case VMCTL_CLEAR_PAGEFAULT:
assert(RTS_ISSET(p,RTS_PAGEFAULT));
RTS_UNSET(p, RTS_PAGEFAULT);
return OK;
case VMCTL_MEMREQ_GET:
/* Send VM the information about the memory request. We can
* not simply send the first request on the list, because IPC
* filters may forbid VM from getting requests for particular
* sources. However, IPC filters are used only in rare cases.
*/
for (rpp = &vmrequest; *rpp != NULL;
rpp = &(*rpp)->p_vmrequest.nextrequestor) {
rp = *rpp;
assert(RTS_ISSET(rp, RTS_VMREQUEST));
okendpt(rp->p_vmrequest.target, &proc_nr);
target = proc_addr(proc_nr);
/* Check against IPC filters. */
if (!allow_ipc_filtered_memreq(rp, target))
continue;
/* Reply with request fields. */
if (rp->p_vmrequest.req_type != VMPTYPE_CHECK)
panic("VMREQUEST wrong type");
m_ptr->SVMCTL_MRG_TARGET =
rp->p_vmrequest.target;
m_ptr->SVMCTL_MRG_ADDR =
rp->p_vmrequest.params.check.start;
m_ptr->SVMCTL_MRG_LENGTH =
rp->p_vmrequest.params.check.length;
m_ptr->SVMCTL_MRG_FLAG =
rp->p_vmrequest.params.check.writeflag;
m_ptr->SVMCTL_MRG_REQUESTOR =
(void *) rp->p_endpoint;
rp->p_vmrequest.vmresult = VMSUSPEND;
/* Remove from request chain. */
*rpp = rp->p_vmrequest.nextrequestor;
return rp->p_vmrequest.req_type;
}
return ENOENT;
case VMCTL_MEMREQ_REPLY:
assert(RTS_ISSET(p, RTS_VMREQUEST));
assert(p->p_vmrequest.vmresult == VMSUSPEND);
okendpt(p->p_vmrequest.target, &proc_nr);
target = proc_addr(proc_nr);
p->p_vmrequest.vmresult = m_ptr->SVMCTL_VALUE;
assert(p->p_vmrequest.vmresult != VMSUSPEND);
switch(p->p_vmrequest.type) {
case VMSTYPE_KERNELCALL:
/*
* we will have to resume execution of the kernel call
* as soon the scheduler picks up this process again
*/
p->p_misc_flags |= MF_KCALL_RESUME;
break;
case VMSTYPE_DELIVERMSG:
assert(p->p_misc_flags & MF_DELIVERMSG);
assert(p == target);
assert(RTS_ISSET(p, RTS_VMREQUEST));
break;
case VMSTYPE_MAP:
assert(RTS_ISSET(p, RTS_VMREQUEST));
break;
default:
panic("strange request type: %d",p->p_vmrequest.type);
}
RTS_UNSET(p, RTS_VMREQUEST);
return OK;
case VMCTL_KERN_PHYSMAP:
{
int i = m_ptr->SVMCTL_VALUE;
return arch_phys_map(i,
(phys_bytes *) &m_ptr->SVMCTL_MAP_PHYS_ADDR,
(phys_bytes *) &m_ptr->SVMCTL_MAP_PHYS_LEN,
&m_ptr->SVMCTL_MAP_FLAGS);
}
case VMCTL_KERN_MAP_REPLY:
{
return arch_phys_map_reply(m_ptr->SVMCTL_VALUE,
(vir_bytes) m_ptr->SVMCTL_MAP_VIR_ADDR);
}
case VMCTL_VMINHIBIT_SET:
/* check if we must stop a process on a different CPU */
#if CONFIG_SMP
if (p->p_cpu != cpuid) {
smp_schedule_vminhibit(p);
} else
#endif
RTS_SET(p, RTS_VMINHIBIT);
#if CONFIG_SMP
p->p_misc_flags |= MF_FLUSH_TLB;
#endif
return OK;
case VMCTL_VMINHIBIT_CLEAR:
assert(RTS_ISSET(p, RTS_VMINHIBIT));
/*
* the processes is certainly not runnable, no need to tell its
* cpu
*/
RTS_UNSET(p, RTS_VMINHIBIT);
#ifdef CONFIG_SMP
if (p->p_misc_flags & MF_SENDA_VM_MISS) {
struct priv *privp;
p->p_misc_flags &= ~MF_SENDA_VM_MISS;
privp = priv(p);
try_deliver_senda(p, (asynmsg_t *) privp->s_asyntab,
privp->s_asynsize);
}
/*
* We don't know whether kernel has the changed mapping
* installed to access userspace memory. And if so, on what CPU.
* More over we don't know what mapping has changed and how and
* therefore we must invalidate all mappings we have anywhere.
* Next time we map memory, we map it fresh.
*/
bits_fill(p->p_stale_tlb, CONFIG_MAX_CPUS);
#endif
return OK;
case VMCTL_CLEARMAPCACHE:
/* VM says: forget about old mappings we have cached. */
mem_clear_mapcache();
return OK;
case VMCTL_BOOTINHIBIT_CLEAR:
RTS_UNSET(p, RTS_BOOTINHIBIT);
return OK;
}
/* Try architecture-specific vmctls. */
return arch_do_vmctl(m_ptr, p);
}
/*===========================================================================*
* do_fork *
*===========================================================================*/
int do_fork(struct proc * caller, message * m_ptr)
{
/* Handle sys_fork().
* m_lsys_krn_sys_fork.endpt has forked.
* The child is m_lsys_krn_sys_fork.slot.
*/
#if defined(__i386__)
char *old_fpu_save_area_p;
#endif
register struct proc *rpc; /* child process pointer */
struct proc *rpp; /* parent process pointer */
int gen;
int p_proc;
int namelen;
if(!isokendpt(m_ptr->m_lsys_krn_sys_fork.endpt, &p_proc))
return EINVAL;
rpp = proc_addr(p_proc);
rpc = proc_addr(m_ptr->m_lsys_krn_sys_fork.slot);
if (isemptyp(rpp) || ! isemptyp(rpc)) return(EINVAL);
assert(!(rpp->p_misc_flags & MF_DELIVERMSG));
/* needs to be receiving so we know where the message buffer is */
if(!RTS_ISSET(rpp, RTS_RECEIVING)) {
printf("kernel: fork not done synchronously?\n");
return EINVAL;
}
/* make sure that the FPU context is saved in parent before copy */
save_fpu(rpp);
/* Copy parent 'proc' struct to child. And reinitialize some fields. */
gen = _ENDPOINT_G(rpc->p_endpoint);
#if defined(__i386__)
old_fpu_save_area_p = rpc->p_seg.fpu_state;
#endif
*rpc = *rpp; /* copy 'proc' struct */
#if defined(__i386__)
rpc->p_seg.fpu_state = old_fpu_save_area_p;
if(proc_used_fpu(rpp))
memcpy(rpc->p_seg.fpu_state, rpp->p_seg.fpu_state, FPU_XFP_SIZE);
#endif
if(++gen >= _ENDPOINT_MAX_GENERATION) /* increase generation */
gen = 1; /* generation number wraparound */
rpc->p_nr = m_ptr->m_lsys_krn_sys_fork.slot; /* this was obliterated by copy */
rpc->p_endpoint = _ENDPOINT(gen, rpc->p_nr); /* new endpoint of slot */
rpc->p_reg.retreg = 0; /* child sees pid = 0 to know it is child */
rpc->p_user_time = 0; /* set all the accounting times to 0 */
rpc->p_sys_time = 0;
rpc->p_misc_flags &=
~(MF_VIRT_TIMER | MF_PROF_TIMER | MF_SC_TRACE | MF_SPROF_SEEN | MF_STEP);
rpc->p_virt_left = 0; /* disable, clear the process-virtual timers */
rpc->p_prof_left = 0;
/* Mark process name as being a forked copy */
namelen = strlen(rpc->p_name);
#define FORKSTR "*F"
if(namelen+strlen(FORKSTR) < sizeof(rpc->p_name))
strcat(rpc->p_name, FORKSTR);
/* the child process is not runnable until it's scheduled. */
RTS_SET(rpc, RTS_NO_QUANTUM);
reset_proc_accounting(rpc);
rpc->p_cpu_time_left = 0;
rpc->p_cycles = 0;
rpc->p_kcall_cycles = 0;
rpc->p_kipc_cycles = 0;
rpc->p_signal_received = 0;
/* If the parent is a privileged process, take away the privileges from the
* child process and inhibit it from running by setting the NO_PRIV flag.
* The caller should explicitly set the new privileges before executing.
*/
if (priv(rpp)->s_flags & SYS_PROC) {
rpc->p_priv = priv_addr(USER_PRIV_ID);
rpc->p_rts_flags |= RTS_NO_PRIV;
}
/* Calculate endpoint identifier, so caller knows what it is. */
m_ptr->m_krn_lsys_sys_fork.endpt = rpc->p_endpoint;
m_ptr->m_krn_lsys_sys_fork.msgaddr = rpp->p_delivermsg_vir;
/* Don't schedule process in VM mode until it has a new pagetable. */
if(m_ptr->m_lsys_krn_sys_fork.flags & PFF_VMINHIBIT) {
RTS_SET(rpc, RTS_VMINHIBIT);
}
/*
* Only one in group should have RTS_SIGNALED, child doesn't inherit tracing.
*/
RTS_UNSET(rpc, (RTS_SIGNALED | RTS_SIG_PENDING | RTS_P_STOP));
(void) sigemptyset(&rpc->p_pending);
#if defined(__i386__)
rpc->p_seg.p_cr3 = 0;
rpc->p_seg.p_cr3_v = NULL;
#elif defined(__arm__)
rpc->p_seg.p_ttbr = 0;
rpc->p_seg.p_ttbr_v = NULL;
#endif
return OK;
}
/*===========================================================================*
* do_privctl *
*===========================================================================*/
PUBLIC int do_privctl(struct proc * caller, message * m_ptr)
{
/* Handle sys_privctl(). Update a process' privileges. If the process is not
* yet a system process, make sure it gets its own privilege structure.
*/
struct proc *rp;
proc_nr_t proc_nr;
sys_id_t priv_id;
int ipc_to_m, kcalls;
int i, r;
struct io_range io_range;
struct mem_range mem_range;
struct priv priv;
int irq;
/* Check whether caller is allowed to make this call. Privileged proceses
* can only update the privileges of processes that are inhibited from
* running by the RTS_NO_PRIV flag. This flag is set when a privileged process
* forks.
*/
if (! (priv(caller)->s_flags & SYS_PROC)) return(EPERM);
if(m_ptr->CTL_ENDPT == SELF) proc_nr = _ENDPOINT_P(caller->p_endpoint);
else if(!isokendpt(m_ptr->CTL_ENDPT, &proc_nr)) return(EINVAL);
rp = proc_addr(proc_nr);
switch(m_ptr->CTL_REQUEST)
{
case SYS_PRIV_ALLOW:
/* Allow process to run. Make sure its privilege structure has already
* been set.
*/
if (!RTS_ISSET(rp, RTS_NO_PRIV) || priv(rp)->s_proc_nr == NONE) {
return(EPERM);
}
RTS_UNSET(rp, RTS_NO_PRIV);
return(OK);
case SYS_PRIV_YIELD:
/* Allow process to run and suspend the caller. */
if (!RTS_ISSET(rp, RTS_NO_PRIV) || priv(rp)->s_proc_nr == NONE) {
return(EPERM);
}
RTS_SET(caller, RTS_NO_PRIV);
RTS_UNSET(rp, RTS_NO_PRIV);
return(OK);
case SYS_PRIV_DISALLOW:
/* Disallow process from running. */
if (RTS_ISSET(rp, RTS_NO_PRIV)) return(EPERM);
RTS_SET(rp, RTS_NO_PRIV);
return(OK);
case SYS_PRIV_SET_SYS:
/* Set a privilege structure of a blocked system process. */
if (! RTS_ISSET(rp, RTS_NO_PRIV)) return(EPERM);
/* Check whether a static or dynamic privilege id must be allocated. */
priv_id = NULL_PRIV_ID;
if (m_ptr->CTL_ARG_PTR)
{
/* Copy privilege structure from caller */
if((r=data_copy(caller->p_endpoint, (vir_bytes) m_ptr->CTL_ARG_PTR,
KERNEL, (vir_bytes) &priv, sizeof(priv))) != OK)
return r;
/* See if the caller wants to assign a static privilege id. */
if(!(priv.s_flags & DYN_PRIV_ID)) {
priv_id = priv.s_id;
}
}
/* Make sure this process has its own privileges structure. This may
* fail, since there are only a limited number of system processes.
* Then copy privileges from the caller and restore some defaults.
*/
if ((i=get_priv(rp, priv_id)) != OK)
{
printf("do_privctl: unable to allocate priv_id %d: %d\n",
priv_id, i);
return(i);
}
priv_id = priv(rp)->s_id; /* backup privilege id */
*priv(rp) = *priv(caller); /* copy from caller */
priv(rp)->s_id = priv_id; /* restore privilege id */
priv(rp)->s_proc_nr = proc_nr; /* reassociate process nr */
for (i=0; i< NR_SYS_CHUNKS; i++) /* remove pending: */
priv(rp)->s_notify_pending.chunk[i] = 0; /* - notifications */
priv(rp)->s_int_pending = 0; /* - interrupts */
(void) sigemptyset(&priv(rp)->s_sig_pending); /* - signals */
reset_timer(&priv(rp)->s_alarm_timer); /* - alarm */
priv(rp)->s_asyntab= -1; /* - asynsends */
priv(rp)->s_asynsize= 0;
/* Set defaults for privilege bitmaps. */
priv(rp)->s_flags= DSRV_F; /* privilege flags */
priv(rp)->s_trap_mask= DSRV_T; /* allowed traps */
ipc_to_m = DSRV_M; /* allowed targets */
fill_sendto_mask(rp, ipc_to_m);
kcalls = DSRV_KC; /* allowed kernel calls */
for(i = 0; i < SYS_CALL_MASK_SIZE; i++) {
priv(rp)->s_k_call_mask[i] = (kcalls == NO_C ? 0 : (~0));
}
/* Set the default signal managers. */
priv(rp)->s_sig_mgr = DSRV_SM;
priv(rp)->s_bak_sig_mgr = NONE;
/* Set defaults for resources: no I/O resources, no memory resources,
* no IRQs, no grant table
*/
priv(rp)->s_nr_io_range= 0;
priv(rp)->s_nr_mem_range= 0;
priv(rp)->s_nr_irq= 0;
priv(rp)->s_grant_table= 0;
priv(rp)->s_grant_entries= 0;
/* Override defaults if the caller has supplied a privilege structure. */
if (m_ptr->CTL_ARG_PTR)
{
if((r = update_priv(rp, &priv)) != OK) {
return r;
}
}
return(OK);
case SYS_PRIV_SET_USER:
/* Set a privilege structure of a blocked user process. */
if (!RTS_ISSET(rp, RTS_NO_PRIV)) return(EPERM);
/* Link the process to the privilege structure of the root user
* process all the user processes share.
*/
priv(rp) = priv_addr(USER_PRIV_ID);
return(OK);
case SYS_PRIV_ADD_IO:
if (RTS_ISSET(rp, RTS_NO_PRIV))
return(EPERM);
/* Only system processes get I/O resources? */
if (!(priv(rp)->s_flags & SYS_PROC))
return EPERM;
#if 0 /* XXX -- do we need a call for this? */
if (strcmp(rp->p_name, "fxp") == 0 ||
strcmp(rp->p_name, "rtl8139") == 0)
{
printf("setting ipc_stats_target to %d\n", rp->p_endpoint);
ipc_stats_target= rp->p_endpoint;
}
#endif
/* Get the I/O range */
data_copy(caller->p_endpoint, (vir_bytes) m_ptr->CTL_ARG_PTR,
KERNEL, (vir_bytes) &io_range, sizeof(io_range));
priv(rp)->s_flags |= CHECK_IO_PORT; /* Check I/O accesses */
i= priv(rp)->s_nr_io_range;
if (i >= NR_IO_RANGE) {
printf("do_privctl: %d already has %d i/o ranges.\n",
rp->p_endpoint, i);
return ENOMEM;
}
priv(rp)->s_io_tab[i].ior_base= io_range.ior_base;
priv(rp)->s_io_tab[i].ior_limit= io_range.ior_limit;
priv(rp)->s_nr_io_range++;
return OK;
case SYS_PRIV_ADD_MEM:
if (RTS_ISSET(rp, RTS_NO_PRIV))
return(EPERM);
/* Only system processes get memory resources? */
if (!(priv(rp)->s_flags & SYS_PROC))
return EPERM;
/* Get the memory range */
if((r=data_copy(caller->p_endpoint, (vir_bytes) m_ptr->CTL_ARG_PTR,
KERNEL, (vir_bytes) &mem_range, sizeof(mem_range))) != OK)
return r;
priv(rp)->s_flags |= CHECK_MEM; /* Check memory mappings */
i= priv(rp)->s_nr_mem_range;
if (i >= NR_MEM_RANGE) {
printf("do_privctl: %d already has %d mem ranges.\n",
rp->p_endpoint, i);
return ENOMEM;
}
priv(rp)->s_mem_tab[i].mr_base= mem_range.mr_base;
priv(rp)->s_mem_tab[i].mr_limit= mem_range.mr_limit;
priv(rp)->s_nr_mem_range++;
return OK;
case SYS_PRIV_ADD_IRQ:
if (RTS_ISSET(rp, RTS_NO_PRIV))
return(EPERM);
/* Only system processes get IRQs? */
if (!(priv(rp)->s_flags & SYS_PROC))
return EPERM;
data_copy(caller->p_endpoint, (vir_bytes) m_ptr->CTL_ARG_PTR,
KERNEL, (vir_bytes) &irq, sizeof(irq));
priv(rp)->s_flags |= CHECK_IRQ; /* Check IRQs */
i= priv(rp)->s_nr_irq;
if (i >= NR_IRQ) {
printf("do_privctl: %d already has %d irq's.\n",
rp->p_endpoint, i);
return ENOMEM;
}
priv(rp)->s_irq_tab[i]= irq;
priv(rp)->s_nr_irq++;
return OK;
case SYS_PRIV_QUERY_MEM:
{
phys_bytes addr, limit;
struct priv *sp;
/* See if a certain process is allowed to map in certain physical
* memory.
*/
addr = (phys_bytes) m_ptr->CTL_PHYSSTART;
limit = addr + (phys_bytes) m_ptr->CTL_PHYSLEN - 1;
if(limit < addr)
return EPERM;
if(!(sp = priv(rp)))
return EPERM;
if (!(sp->s_flags & SYS_PROC))
return EPERM;
for(i = 0; i < sp->s_nr_mem_range; i++) {
if(addr >= sp->s_mem_tab[i].mr_base &&
limit <= sp->s_mem_tab[i].mr_limit)
return OK;
}
return EPERM;
}
case SYS_PRIV_UPDATE_SYS:
/* Update the privilege structure of a system process. */
if(!m_ptr->CTL_ARG_PTR) return EINVAL;
/* Copy privilege structure from caller */
if((r=data_copy(caller->p_endpoint, (vir_bytes) m_ptr->CTL_ARG_PTR,
KERNEL, (vir_bytes) &priv, sizeof(priv))) != OK)
return r;
/* Override settings in existing privilege structure. */
if((r = update_priv(rp, &priv)) != OK) {
return r;
}
return(OK);
default:
printf("do_privctl: bad request %d\n", m_ptr->CTL_REQUEST);
return EINVAL;
}
}
/*===========================================================================*
* kmain *
*===========================================================================*/
void kmain(kinfo_t *local_cbi)
{
/* Start the ball rolling. */
struct boot_image *ip; /* boot image pointer */
register struct proc *rp; /* process pointer */
register int i, j;
/* save a global copy of the boot parameters */
memcpy(&kinfo, local_cbi, sizeof(kinfo));
memcpy(&kmess, kinfo.kmess, sizeof(kmess));
#ifdef __arm__
/* We want to initialize serial before we do any output */
omap3_ser_init();
#endif
/* We can talk now */
printf("MINIX booting\n");
/* Kernel may use bits of main memory before VM is started */
kernel_may_alloc = 1;
assert(sizeof(kinfo.boot_procs) == sizeof(image));
memcpy(kinfo.boot_procs, image, sizeof(kinfo.boot_procs));
cstart();
BKL_LOCK();
DEBUGEXTRA(("main()\n"));
proc_init();
if(NR_BOOT_MODULES != kinfo.mbi.mods_count)
panic("expecting %d boot processes/modules, found %d",
NR_BOOT_MODULES, kinfo.mbi.mods_count);
/* Set up proc table entries for processes in boot image. */
for (i=0; i < NR_BOOT_PROCS; ++i) {
int schedulable_proc;
proc_nr_t proc_nr;
int ipc_to_m, kcalls;
sys_map_t map;
ip = &image[i]; /* process' attributes */
DEBUGEXTRA(("initializing %s... ", ip->proc_name));
rp = proc_addr(ip->proc_nr); /* get process pointer */
ip->endpoint = rp->p_endpoint; /* ipc endpoint */
make_zero64(rp->p_cpu_time_left);
if(i < NR_TASKS) /* name (tasks only) */
strlcpy(rp->p_name, ip->proc_name, sizeof(rp->p_name));
if(i >= NR_TASKS) {
/* Remember this so it can be passed to VM */
multiboot_module_t *mb_mod = &kinfo.module_list[i - NR_TASKS];
ip->start_addr = mb_mod->mod_start;
ip->len = mb_mod->mod_end - mb_mod->mod_start;
}
reset_proc_accounting(rp);
/* See if this process is immediately schedulable.
* In that case, set its privileges now and allow it to run.
* Only kernel tasks and the root system process get to run immediately.
* All the other system processes are inhibited from running by the
* RTS_NO_PRIV flag. They can only be scheduled once the root system
* process has set their privileges.
*/
proc_nr = proc_nr(rp);
schedulable_proc = (iskerneln(proc_nr) || isrootsysn(proc_nr) ||
proc_nr == VM_PROC_NR);
if(schedulable_proc) {
/* Assign privilege structure. Force a static privilege id. */
(void) get_priv(rp, static_priv_id(proc_nr));
/* Priviliges for kernel tasks. */
if(proc_nr == VM_PROC_NR) {
priv(rp)->s_flags = VM_F;
priv(rp)->s_trap_mask = SRV_T;
ipc_to_m = SRV_M;
kcalls = SRV_KC;
priv(rp)->s_sig_mgr = SELF;
rp->p_priority = SRV_Q;
rp->p_quantum_size_ms = SRV_QT;
}
else if(iskerneln(proc_nr)) {
/* Privilege flags. */
priv(rp)->s_flags = (proc_nr == IDLE ? IDL_F : TSK_F);
/* Allowed traps. */
priv(rp)->s_trap_mask = (proc_nr == CLOCK
|| proc_nr == SYSTEM ? CSK_T : TSK_T);
ipc_to_m = TSK_M; /* allowed targets */
kcalls = TSK_KC; /* allowed kernel calls */
}
/* Priviliges for the root system process. */
else {
assert(isrootsysn(proc_nr));
priv(rp)->s_flags= RSYS_F; /* privilege flags */
priv(rp)->s_trap_mask= SRV_T; /* allowed traps */
ipc_to_m = SRV_M; /* allowed targets */
kcalls = SRV_KC; /* allowed kernel calls */
priv(rp)->s_sig_mgr = SRV_SM; /* signal manager */
rp->p_priority = SRV_Q; /* priority queue */
rp->p_quantum_size_ms = SRV_QT; /* quantum size */
}
/* Fill in target mask. */
memset(&map, 0, sizeof(map));
if (ipc_to_m == ALL_M) {
for(j = 0; j < NR_SYS_PROCS; j++)
set_sys_bit(map, j);
}
fill_sendto_mask(rp, &map);
/* Fill in kernel call mask. */
for(j = 0; j < SYS_CALL_MASK_SIZE; j++) {
priv(rp)->s_k_call_mask[j] = (kcalls == NO_C ? 0 : (~0));
}
}
else {
/* Don't let the process run for now. */
RTS_SET(rp, RTS_NO_PRIV | RTS_NO_QUANTUM);
}
/* Arch-specific state initialization. */
arch_boot_proc(ip, rp);
/* scheduling functions depend on proc_ptr pointing somewhere. */
if(!get_cpulocal_var(proc_ptr))
get_cpulocal_var(proc_ptr) = rp;
/* Process isn't scheduled until VM has set up a pagetable for it. */
if(rp->p_nr != VM_PROC_NR && rp->p_nr >= 0) {
rp->p_rts_flags |= RTS_VMINHIBIT;
rp->p_rts_flags |= RTS_BOOTINHIBIT;
}
rp->p_rts_flags |= RTS_PROC_STOP;
rp->p_rts_flags &= ~RTS_SLOT_FREE;
DEBUGEXTRA(("done\n"));
}
/* update boot procs info for VM */
memcpy(kinfo.boot_procs, image, sizeof(kinfo.boot_procs));
#define IPCNAME(n) { \
assert((n) >= 0 && (n) <= IPCNO_HIGHEST); \
assert(!ipc_call_names[n]); \
ipc_call_names[n] = #n; \
}
arch_post_init();
IPCNAME(SEND);
IPCNAME(RECEIVE);
IPCNAME(SENDREC);
IPCNAME(NOTIFY);
IPCNAME(SENDNB);
IPCNAME(SENDA);
/* System and processes initialization */
memory_init();
DEBUGEXTRA(("system_init()... "));
system_init();
DEBUGEXTRA(("done\n"));
/* The bootstrap phase is over, so we can add the physical
* memory used for it to the free list.
*/
add_memmap(&kinfo, kinfo.bootstrap_start, kinfo.bootstrap_len);
#ifdef CONFIG_SMP
if (config_no_apic) {
BOOT_VERBOSE(printf("APIC disabled, disables SMP, using legacy PIC\n"));
smp_single_cpu_fallback();
} else if (config_no_smp) {
BOOT_VERBOSE(printf("SMP disabled, using legacy PIC\n"));
smp_single_cpu_fallback();
} else {
smp_init();
/*
* if smp_init() returns it means that it failed and we try to finish
* single CPU booting
*/
bsp_finish_booting();
}
#else
/*
* if configured for a single CPU, we are already on the kernel stack which we
* are going to use everytime we execute kernel code. We finish booting and we
* never return here
*/
bsp_finish_booting();
#endif
NOT_REACHABLE;
}
/*===========================================================================*
* do_vmctl *
*===========================================================================*/
PUBLIC int do_vmctl(struct proc * caller, message * m_ptr)
{
int proc_nr;
endpoint_t ep = m_ptr->SVMCTL_WHO;
struct proc *p, *rp, *target;
int err;
if(ep == SELF) { ep = caller->p_endpoint; }
if(!isokendpt(ep, &proc_nr)) {
printf("do_vmctl: unexpected endpoint %d from VM\n", ep);
return EINVAL;
}
p = proc_addr(proc_nr);
switch(m_ptr->SVMCTL_PARAM) {
case VMCTL_CLEAR_PAGEFAULT:
assert(RTS_ISSET(p,RTS_PAGEFAULT));
RTS_UNSET(p, RTS_PAGEFAULT);
return OK;
case VMCTL_MEMREQ_GET:
/* Send VM the information about the memory request. */
if(!(rp = vmrequest))
return ESRCH;
assert(RTS_ISSET(rp, RTS_VMREQUEST));
okendpt(rp->p_vmrequest.target, &proc_nr);
target = proc_addr(proc_nr);
/* Reply with request fields. */
switch(rp->p_vmrequest.req_type) {
case VMPTYPE_CHECK:
m_ptr->SVMCTL_MRG_TARGET =
rp->p_vmrequest.target;
m_ptr->SVMCTL_MRG_ADDR =
rp->p_vmrequest.params.check.start;
m_ptr->SVMCTL_MRG_LENGTH =
rp->p_vmrequest.params.check.length;
m_ptr->SVMCTL_MRG_FLAG =
rp->p_vmrequest.params.check.writeflag;
m_ptr->SVMCTL_MRG_REQUESTOR =
(void *) rp->p_endpoint;
break;
case VMPTYPE_SMAP:
case VMPTYPE_SUNMAP:
case VMPTYPE_COWMAP:
assert(RTS_ISSET(target,RTS_VMREQTARGET));
RTS_UNSET(target, RTS_VMREQTARGET);
m_ptr->SVMCTL_MRG_TARGET =
rp->p_vmrequest.target;
m_ptr->SVMCTL_MRG_ADDR =
rp->p_vmrequest.params.map.vir_d;
m_ptr->SVMCTL_MRG_EP2 =
rp->p_vmrequest.params.map.ep_s;
m_ptr->SVMCTL_MRG_ADDR2 =
rp->p_vmrequest.params.map.vir_s;
m_ptr->SVMCTL_MRG_LENGTH =
rp->p_vmrequest.params.map.length;
m_ptr->SVMCTL_MRG_FLAG =
rp->p_vmrequest.params.map.writeflag;
m_ptr->SVMCTL_MRG_REQUESTOR =
(void *) rp->p_endpoint;
break;
default:
panic("VMREQUEST wrong type");
}
rp->p_vmrequest.vmresult = VMSUSPEND;
/* Remove from request chain. */
vmrequest = vmrequest->p_vmrequest.nextrequestor;
return rp->p_vmrequest.req_type;
case VMCTL_MEMREQ_REPLY:
assert(RTS_ISSET(p, RTS_VMREQUEST));
assert(p->p_vmrequest.vmresult == VMSUSPEND);
okendpt(p->p_vmrequest.target, &proc_nr);
target = proc_addr(proc_nr);
p->p_vmrequest.vmresult = m_ptr->SVMCTL_VALUE;
assert(p->p_vmrequest.vmresult != VMSUSPEND);
switch(p->p_vmrequest.type) {
case VMSTYPE_KERNELCALL:
/*
* we will have to resume execution of the kernel call
* as soon the scheduler picks up this process again
*/
p->p_misc_flags |= MF_KCALL_RESUME;
break;
case VMSTYPE_DELIVERMSG:
assert(p->p_misc_flags & MF_DELIVERMSG);
assert(p == target);
assert(RTS_ISSET(p, RTS_VMREQUEST));
break;
case VMSTYPE_MAP:
assert(RTS_ISSET(p, RTS_VMREQUEST));
break;
default:
panic("strange request type: %d",p->p_vmrequest.type);
}
RTS_UNSET(p, RTS_VMREQUEST);
return OK;
case VMCTL_ENABLE_PAGING:
if(vm_running)
panic("do_vmctl: paging already enabled");
if (arch_enable_paging(caller, m_ptr) != OK)
panic("do_vmctl: paging enabling failed");
return OK;
case VMCTL_KERN_PHYSMAP:
{
int i = m_ptr->SVMCTL_VALUE;
return arch_phys_map(i,
(phys_bytes *) &m_ptr->SVMCTL_MAP_PHYS_ADDR,
(phys_bytes *) &m_ptr->SVMCTL_MAP_PHYS_LEN,
&m_ptr->SVMCTL_MAP_FLAGS);
}
case VMCTL_KERN_MAP_REPLY:
{
return arch_phys_map_reply(m_ptr->SVMCTL_VALUE,
(vir_bytes) m_ptr->SVMCTL_MAP_VIR_ADDR);
}
case VMCTL_VMINHIBIT_SET:
/* check if we must stop a process on a different CPU */
#if CONFIG_SMP
if (p->p_cpu != cpuid) {
smp_schedule_vminhibit(p);
} else
#endif
RTS_SET(p, RTS_VMINHIBIT);
#if CONFIG_SMP
p->p_misc_flags |= MF_FLUSH_TLB;
#endif
return OK;
case VMCTL_VMINHIBIT_CLEAR:
assert(RTS_ISSET(p, RTS_VMINHIBIT));
/*
* the processes is certainly not runnable, no need to tell its
* cpu
*/
RTS_UNSET(p, RTS_VMINHIBIT);
return OK;
}
/* Try architecture-specific vmctls. */
return arch_do_vmctl(m_ptr, p);
}
/*===========================================================================*
* main *
*===========================================================================*/
PUBLIC int main(void)
{
/* Start the ball rolling. */
struct boot_image *ip; /* boot image pointer */
register struct proc *rp; /* process pointer */
register int i, j;
size_t argsz; /* size of arguments passed to crtso on stack */
BKL_LOCK();
/* Global value to test segment sanity. */
magictest = MAGICTEST;
DEBUGEXTRA(("main()\n"));
proc_init();
/* Set up proc table entries for processes in boot image. The stacks
* of the servers have been added to the data segment by the monitor, so
* the stack pointer is set to the end of the data segment.
*/
for (i=0; i < NR_BOOT_PROCS; ++i) {
int schedulable_proc;
proc_nr_t proc_nr;
int ipc_to_m, kcalls;
sys_map_t map;
ip = &image[i]; /* process' attributes */
DEBUGEXTRA(("initializing %s... ", ip->proc_name));
rp = proc_addr(ip->proc_nr); /* get process pointer */
ip->endpoint = rp->p_endpoint; /* ipc endpoint */
make_zero64(rp->p_cpu_time_left);
strncpy(rp->p_name, ip->proc_name, P_NAME_LEN); /* set process name */
reset_proc_accounting(rp);
/* See if this process is immediately schedulable.
* In that case, set its privileges now and allow it to run.
* Only kernel tasks and the root system process get to run immediately.
* All the other system processes are inhibited from running by the
* RTS_NO_PRIV flag. They can only be scheduled once the root system
* process has set their privileges.
*/
proc_nr = proc_nr(rp);
schedulable_proc = (iskerneln(proc_nr) || isrootsysn(proc_nr));
if(schedulable_proc) {
/* Assign privilege structure. Force a static privilege id. */
(void) get_priv(rp, static_priv_id(proc_nr));
/* Priviliges for kernel tasks. */
if(iskerneln(proc_nr)) {
/* Privilege flags. */
priv(rp)->s_flags = (proc_nr == IDLE ? IDL_F : TSK_F);
/* Allowed traps. */
priv(rp)->s_trap_mask = (proc_nr == CLOCK
|| proc_nr == SYSTEM ? CSK_T : TSK_T);
ipc_to_m = TSK_M; /* allowed targets */
kcalls = TSK_KC; /* allowed kernel calls */
}
/* Priviliges for the root system process. */
else if(isrootsysn(proc_nr)) {
priv(rp)->s_flags= RSYS_F; /* privilege flags */
priv(rp)->s_trap_mask= SRV_T; /* allowed traps */
ipc_to_m = SRV_M; /* allowed targets */
kcalls = SRV_KC; /* allowed kernel calls */
priv(rp)->s_sig_mgr = SRV_SM; /* signal manager */
rp->p_priority = SRV_Q; /* priority queue */
rp->p_quantum_size_ms = SRV_QT; /* quantum size */
}
/* Priviliges for ordinary process. */
else {
NOT_REACHABLE;
}
/* Fill in target mask. */
memset(&map, 0, sizeof(map));
if (ipc_to_m == ALL_M) {
for(j = 0; j < NR_SYS_PROCS; j++)
set_sys_bit(map, j);
}
fill_sendto_mask(rp, &map);
/* Fill in kernel call mask. */
for(j = 0; j < SYS_CALL_MASK_SIZE; j++) {
priv(rp)->s_k_call_mask[j] = (kcalls == NO_C ? 0 : (~0));
}
}
else {
/* Don't let the process run for now. */
RTS_SET(rp, RTS_NO_PRIV | RTS_NO_QUANTUM);
}
rp->p_memmap[T].mem_vir = ABS2CLICK(ip->memmap.text_vaddr);
rp->p_memmap[T].mem_phys = ABS2CLICK(ip->memmap.text_paddr);
rp->p_memmap[T].mem_len = ABS2CLICK(ip->memmap.text_bytes);
rp->p_memmap[D].mem_vir = ABS2CLICK(ip->memmap.data_vaddr);
rp->p_memmap[D].mem_phys = ABS2CLICK(ip->memmap.data_paddr);
rp->p_memmap[D].mem_len = ABS2CLICK(ip->memmap.data_bytes);
rp->p_memmap[S].mem_phys = ABS2CLICK(ip->memmap.data_paddr +
ip->memmap.data_bytes +
ip->memmap.stack_bytes);
rp->p_memmap[S].mem_vir = ABS2CLICK(ip->memmap.data_vaddr +
ip->memmap.data_bytes +
ip->memmap.stack_bytes);
rp->p_memmap[S].mem_len = 0;
/* Set initial register values. The processor status word for tasks
* is different from that of other processes because tasks can
* access I/O; this is not allowed to less-privileged processes
*/
rp->p_reg.pc = ip->memmap.entry;
rp->p_reg.psw = (iskerneln(proc_nr)) ? INIT_TASK_PSW : INIT_PSW;
/* Initialize the server stack pointer. Take it down three words
* to give crtso.s something to use as "argc", "argv" and "envp".
*/
if (isusern(proc_nr)) { /* user-space process? */
rp->p_reg.sp = (rp->p_memmap[S].mem_vir +
rp->p_memmap[S].mem_len) << CLICK_SHIFT;
argsz = 3 * sizeof(reg_t);
rp->p_reg.sp -= argsz;
phys_memset(rp->p_reg.sp -
(rp->p_memmap[S].mem_vir << CLICK_SHIFT) +
(rp->p_memmap[S].mem_phys << CLICK_SHIFT),
0, argsz);
}
/* scheduling functions depend on proc_ptr pointing somewhere. */
if(!get_cpulocal_var(proc_ptr))
get_cpulocal_var(proc_ptr) = rp;
/* If this process has its own page table, VM will set the
* PT up and manage it. VM will signal the kernel when it has
* done this; until then, don't let it run.
*/
if(ip->flags & PROC_FULLVM)
rp->p_rts_flags |= RTS_VMINHIBIT;
rp->p_rts_flags |= RTS_PROC_STOP;
rp->p_rts_flags &= ~RTS_SLOT_FREE;
alloc_segments(rp);
DEBUGEXTRA(("done\n"));
}
#define IPCNAME(n) { \
assert((n) >= 0 && (n) <= IPCNO_HIGHEST); \
assert(!ipc_call_names[n]); \
ipc_call_names[n] = #n; \
}
IPCNAME(SEND);
IPCNAME(RECEIVE);
IPCNAME(SENDREC);
IPCNAME(NOTIFY);
IPCNAME(SENDNB);
IPCNAME(SENDA);
/* Architecture-dependent initialization. */
DEBUGEXTRA(("arch_init()... "));
arch_init();
DEBUGEXTRA(("done\n"));
/* System and processes initialization */
DEBUGEXTRA(("system_init()... "));
system_init();
DEBUGEXTRA(("done\n"));
#ifdef CONFIG_SMP
if (config_no_apic) {
BOOT_VERBOSE(printf("APIC disabled, disables SMP, using legacy PIC\n"));
smp_single_cpu_fallback();
} else if (config_no_smp) {
BOOT_VERBOSE(printf("SMP disabled, using legacy PIC\n"));
smp_single_cpu_fallback();
} else {
smp_init();
/*
* if smp_init() returns it means that it failed and we try to finish
* single CPU booting
*/
bsp_finish_booting();
}
#else
/*
* if configured for a single CPU, we are already on the kernel stack which we
* are going to use everytime we execute kernel code. We finish booting and we
* never return here
*/
bsp_finish_booting();
#endif
NOT_REACHABLE;
return 1;
}
void main(void)
{
/* Start the ball rolling. */
struct boot_image *ip; /* boot image pointer */
register struct proc *rp; /* process pointer */
register struct priv *sp; /* privilege structure pointer */
register int i, j;
int hdrindex; /* index to array of a.out headers */
phys_clicks text_base;
vir_clicks text_clicks, data_clicks, st_clicks;
reg_t ktsb; /* kernel task stack base */
struct exec *e_hdr = 0; /* for a copy of an a.out header */
/* Global value to test segment sanity. */
magictest = MAGICTEST;
/* Clear the process table. Anounce each slot as empty and set up mappings
* for proc_addr() and proc_nr() macros. Do the same for the table with
* privilege structures for the system processes.
*/
for (rp = BEG_PROC_ADDR, i = -NR_TASKS; rp < END_PROC_ADDR; ++rp, ++i) {
rp->p_rts_flags = RTS_SLOT_FREE; /* initialize free slot */
#ifdef CONFIG_DEBUG_KERNEL_SCHED_CHECK
rp->p_magic = PMAGIC;
#endif
rp->p_nr = i; /* proc number from ptr */
rp->p_endpoint = _ENDPOINT(0, rp->p_nr); /* generation no. 0 */
}
for (sp = BEG_PRIV_ADDR, i = 0; sp < END_PRIV_ADDR; ++sp, ++i) {
sp->s_proc_nr = ENDPT_NONE; /* initialize as free */
sp->s_id = i; /* priv structure index */
ppriv_addr[i] = sp; /* priv ptr from number */
}
/* Set up proc table entries for processes in boot image. The stacks of the
* kernel tasks are initialized to an array in data space. The stacks
* of the servers have been added to the data segment by the monitor, so
* the stack pointer is set to the end of the data segment. All the
* processes are in low memory on the 8086. On the 386 only the kernel
* is in low memory, the rest is loaded in extended memory.
*/
/* Task stacks. */
ktsb = (reg_t) t_stack;
for (i=0; i < NR_BOOT_PROCS; ++i) {
int schedulable_proc, proc_nr;
int ipc_to_m, kcalls;
ip = &image[i]; /* process' attributes */
rp = proc_addr(ip->proc_nr); /* get process pointer */
ip->endpoint = rp->p_endpoint; /* ipc endpoint */
rp->p_max_priority = ip->priority; /* max scheduling priority */
rp->p_priority = ip->priority; /* current priority */
rp->p_quantum_size = ip->quantum; /* quantum size in ticks */
rp->p_ticks_left = ip->quantum; /* current credit */
strncpy(rp->p_name, ip->proc_name, P_NAME_LEN); /* set process name */
/* See if this process is immediately schedulable.
* In that case, set its privileges now and allow it to run.
* Only kernel tasks and the root system process get to run immediately.
* All the other system processes are inhibited from running by the
* RTS_NO_PRIV flag. They can only be scheduled once the root system
* process has set their privileges.
*/
proc_nr = proc_nr(rp);
schedulable_proc = (iskerneln(proc_nr) || isrootsysn(proc_nr));
if(schedulable_proc) {
/* Assign privilege structure. Force a static privilege id. */
(void) get_priv(rp, static_priv_id(proc_nr));
/* Priviliges for kernel tasks. */
if(iskerneln(proc_nr)) {
/* Privilege flags. */
priv(rp)->s_flags = (proc_nr == IDLE ? IDL_F : TSK_F);
/* Allowed traps. */
priv(rp)->s_trap_mask = (proc_nr == CLOCK
|| proc_nr == SYSTEM ? CSK_T : TSK_T);
ipc_to_m = TSK_M; /* allowed targets */
kcalls = TSK_KC; /* allowed kernel calls */
} else if(isrootsysn(proc_nr)) {
/* Priviliges for the root system process. */
priv(rp)->s_flags= RSYS_F; /* privilege flags */
priv(rp)->s_trap_mask= RSYS_T; /* allowed traps */
ipc_to_m = RSYS_M; /* allowed targets */
kcalls = RSYS_KC; /* allowed kernel calls */
}
/* Fill in target mask. */
for (j=0; j < NR_SYS_PROCS; j++) {
if (ipc_to_m & (1 << j))
set_sendto_bit(rp, j);
else
unset_sendto_bit(rp, j);
}
/* Fill in kernel call mask. */
for(j = 0; j < CALL_MASK_SIZE; j++) {
priv(rp)->s_k_call_mask[j] = (kcalls == NO_C ? 0 : (~0));
}
} else {
/*Don't let the process run for now. */
RTS_SET(rp, RTS_NO_PRIV);
}
if (iskerneln(proc_nr)) { /* part of the kernel? */
if (ip->stksize > 0) { /* HARDWARE stack size is 0 */
rp->p_priv->s_stack_guard = (reg_t *) ktsb;
*rp->p_priv->s_stack_guard = STACK_GUARD;
}
ktsb += ip->stksize; /* point to high end of stack */
rp->p_reg.sp = ktsb; /* this task's initial stack ptr */
hdrindex = 0; /* all use the first a.out header */
} else {
hdrindex = 1 + i-NR_TASKS; /* system/user processes */
}
/* Architecture-specific way to find out aout header of this
* boot process.
*/
e_hdr = arch_get_aout_header(hdrindex);
/* Convert addresses to clicks and build process memory map */
text_base = e_hdr->a_syms >> CLICK_SHIFT;
st_clicks= (e_hdr->a_total + CLICK_SIZE-1) >> CLICK_SHIFT;
data_clicks = (e_hdr->a_text + e_hdr->a_data + e_hdr->a_bss + CLICK_SIZE-1) >> CLICK_SHIFT;
text_clicks = 0;
rp->p_memmap[T].mem_phys = text_base;
rp->p_memmap[T].mem_len = text_clicks;
rp->p_memmap[D].mem_phys = text_base + text_clicks;
rp->p_memmap[D].mem_len = data_clicks;
rp->p_memmap[S].mem_phys = text_base + text_clicks + st_clicks;
rp->p_memmap[S].mem_vir = st_clicks;
rp->p_memmap[S].mem_len = 0;
/* Patch (override) the non-kernel process' entry points in image table. The
* image table is located in kernel/kernel_syms.c. The kernel processes like
* IDLE, SYSTEM, CLOCK, HARDWARE are not changed because they are part of kernel
* and the entry points are set at compilation time. In case of IDLE or HARDWARE
* the entry point can be ignored becasue they never run (set RTS_PROC_STOP).
*/
if (!iskerneln(proc_nr(rp)))
ip->initial_pc = (task_t*)e_hdr->a_entry;
/* Set initial register values. The processor status word for tasks
* is different from that of other processes because tasks can
* access I/O; this is not allowed to less-privileged processes
*/
rp->p_reg.pc = (reg_t) ip->initial_pc;
rp->p_reg.psw = (iskerneln(proc_nr)) ? INIT_TASK_PSW : INIT_PSW;
/* Initialize the server stack pointer. Take it down one word
* to give crtso.s something to use as "argc","argv" and "envp".
*/
if (isusern(proc_nr)) { /* user-space process? */
rp->p_reg.sp = (rp->p_memmap[S].mem_vir + rp->p_memmap[S].mem_len)
<< CLICK_SHIFT;
rp->p_reg.sp -= 3*sizeof(reg_t);
}
/* scheduling functions depend on proc_ptr pointing somewhere. */
if(!proc_ptr)
proc_ptr = rp;
/* If this process has its own page table, VM will set the
* PT up and manage it. VM will signal the kernel when it has
* done this; until then, don't let it run.
*/
if(ip->flags & PROC_FULLVM)
RTS_SET(rp, RTS_VMINHIBIT);
/* IDLE & HARDWARE task is never put on a run queue as it is
* never ready to run.
*/
if (rp->p_nr == HARDWARE)
RTS_SET(rp, RTS_PROC_STOP);
if (rp->p_nr == IDLE)
RTS_SET(rp, RTS_PROC_STOP);
RTS_UNSET(rp, RTS_SLOT_FREE); /* remove RTS_SLOT_FREE and schedule */
alloc_segments(rp);
} /* for */
/* Architecture-dependent initialization. */
arch_init();
#ifdef CONFIG_DEBUG_KERNEL_STATS_PROFILE
sprofiling = 0; /* we're not profiling until instructed to */
#endif
cprof_procs_no = 0; /* init nr of hash table slots used */
#ifdef CONFIG_IDLE_TSC
idle_tsc = cvu64(0);
#endif
vm_running = 0;
krandom.random_sources = RANDOM_SOURCES;
krandom.random_elements = RANDOM_ELEMENTS;
/* Nucleos is now ready. All boot image processes are on the ready queue.
* Return to the assembly code to start running the current process.
*/
bill_ptr = proc_addr(IDLE); /* it has to point somewhere */
announce(); /* print Nucleos startup banner */
/*
* enable timer interrupts and clock task on the boot CPU
*/
if (boot_cpu_init_timer(system_hz)) {
kernel_panic("FATAL : failed to initialize timer interrupts, "
"cannot continue without any clock source!",
NO_NUM);
}
/* Warnings for sanity checks that take time. These warnings are printed
* so it's a clear warning no full release should be done with them
* enabled.
*/
#ifdef CONFIG_DEBUG_KERNEL_SCHED_CHECK
FIXME("CONFIG_DEBUG_KERNEL_SCHED_CHECK enabled");
#endif
#ifdef CONFIG_DEBUG_KERNEL_VMASSERT
FIXME("CONFIG_DEBUG_KERNEL_VMASSERT enabled");
#endif
#ifdef CONFIG_DEBUG_PROC_CHECK
FIXME("PROC check enabled");
#endif
restart();
}
PRIVATE void pagefault( struct proc *pr,
struct exception_frame * frame,
int is_nested)
{
int in_physcopy = 0;
reg_t pagefaultcr2;
message m_pagefault;
int err;
assert(frame);
pagefaultcr2 = read_cr2();
#if 0
printf("kernel: pagefault in pr %d, addr 0x%lx, his cr3 0x%lx, actual cr3 0x%lx\n",
pr->p_endpoint, pagefaultcr2, pr->p_seg.p_cr3, read_cr3());
#endif
if(pr->p_seg.p_cr3) {
assert(pr->p_seg.p_cr3 == read_cr3());
}
in_physcopy = (frame->eip > (vir_bytes) phys_copy) &&
(frame->eip < (vir_bytes) phys_copy_fault);
if((is_nested || iskernelp(pr)) &&
catch_pagefaults && in_physcopy) {
#if 0
printf("pf caught! addr 0x%lx\n", pagefaultcr2);
#endif
if (is_nested) {
frame->eip = (reg_t) phys_copy_fault_in_kernel;
}
else {
pr->p_reg.pc = (reg_t) phys_copy_fault;
pr->p_reg.retreg = pagefaultcr2;
}
return;
}
if(is_nested) {
panic("pagefault in kernel at pc 0x%lx address 0x%lx", frame->eip, pagefaultcr2);
}
/* System processes that don't have their own page table can't
* have page faults. VM does have its own page table but also
* can't have page faults (because VM has to handle them).
*/
if((pr->p_endpoint <= INIT_PROC_NR &&
!(pr->p_misc_flags & MF_FULLVM)) || pr->p_endpoint == VM_PROC_NR) {
/* Page fault we can't / don't want to
* handle.
*/
printf("pagefault for process %d ('%s'), pc = 0x%x, addr = 0x%x, flags = 0x%x, is_nested %d\n",
pr->p_endpoint, pr->p_name, pr->p_reg.pc,
pagefaultcr2, frame->errcode, is_nested);
proc_stacktrace(pr);
printf("pc of pagefault: 0x%lx\n", frame->eip);
panic("page fault in system process: %d", pr->p_endpoint);
return;
}
/* Don't schedule this process until pagefault is handled. */
assert(pr->p_seg.p_cr3 == read_cr3());
assert(!RTS_ISSET(pr, RTS_PAGEFAULT));
RTS_SET(pr, RTS_PAGEFAULT);
/* tell Vm about the pagefault */
m_pagefault.m_source = pr->p_endpoint;
m_pagefault.m_type = VM_PAGEFAULT;
m_pagefault.VPF_ADDR = pagefaultcr2;
m_pagefault.VPF_FLAGS = frame->errcode;
if ((err = mini_send(pr, VM_PROC_NR,
&m_pagefault, FROM_KERNEL))) {
panic("WARNING: pagefault: mini_send returned %d\n", err);
}
return;
}
/*===========================================================================*
* main *
*===========================================================================*/
PUBLIC void main()
{
/* Start the ball rolling. */
struct boot_image *ip; /* boot image pointer */
register struct proc *rp; /* process pointer */
register struct priv *sp; /* privilege structure pointer */
register int i, j, s;
int hdrindex; /* index to array of a.out headers */
phys_clicks text_base;
vir_clicks text_clicks, data_clicks, st_clicks;
reg_t ktsb; /* kernel task stack base */
struct exec e_hdr; /* for a copy of an a.out header */
/* Architecture-dependent initialization. */
arch_init();
/* Global value to test segment sanity. */
magictest = MAGICTEST;
/* Clear the process table. Anounce each slot as empty and set up mappings
* for proc_addr() and proc_nr() macros. Do the same for the table with
* privilege structures for the system processes.
*/
for (rp = BEG_PROC_ADDR, i = -NR_TASKS; rp < END_PROC_ADDR; ++rp, ++i) {
rp->p_rts_flags = SLOT_FREE; /* initialize free slot */
#if DEBUG_SCHED_CHECK
rp->p_magic = PMAGIC;
#endif
rp->p_nr = i; /* proc number from ptr */
rp->p_endpoint = _ENDPOINT(0, rp->p_nr); /* generation no. 0 */
}
for (sp = BEG_PRIV_ADDR, i = 0; sp < END_PRIV_ADDR; ++sp, ++i) {
sp->s_proc_nr = NONE; /* initialize as free */
sp->s_id = i; /* priv structure index */
ppriv_addr[i] = sp; /* priv ptr from number */
}
/* Set up proc table entries for processes in boot image. The stacks of the
* kernel tasks are initialized to an array in data space. The stacks
* of the servers have been added to the data segment by the monitor, so
* the stack pointer is set to the end of the data segment. All the
* processes are in low memory on the 8086. On the 386 only the kernel
* is in low memory, the rest is loaded in extended memory.
*/
/* Task stacks. */
ktsb = (reg_t) t_stack;
for (i=0; i < NR_BOOT_PROCS; ++i) {
int ci;
bitchunk_t fv;
ip = &image[i]; /* process' attributes */
rp = proc_addr(ip->proc_nr); /* get process pointer */
ip->endpoint = rp->p_endpoint; /* ipc endpoint */
rp->p_max_priority = ip->priority; /* max scheduling priority */
rp->p_priority = ip->priority; /* current priority */
rp->p_quantum_size = ip->quantum; /* quantum size in ticks */
rp->p_ticks_left = ip->quantum; /* current credit */
strncpy(rp->p_name, ip->proc_name, P_NAME_LEN); /* set process name */
(void) get_priv(rp, (ip->flags & SYS_PROC)); /* assign structure */
priv(rp)->s_flags = ip->flags; /* process flags */
priv(rp)->s_trap_mask = ip->trap_mask; /* allowed traps */
/* Warn about violations of the boot image table order consistency. */
if (priv_id(rp) != s_nr_to_id(ip->proc_nr) && (ip->flags & SYS_PROC))
kprintf("Warning: boot image table has wrong process order\n");
/* Initialize call mask bitmap from unordered set.
* A single SYS_ALL_CALLS is a special case - it
* means all calls are allowed.
*/
if(ip->nr_k_calls == 1 && ip->k_calls[0] == SYS_ALL_CALLS)
fv = ~0; /* fill call mask */
else
fv = 0; /* clear call mask */
for(ci = 0; ci < CALL_MASK_SIZE; ci++) /* fill or clear call mask */
priv(rp)->s_k_call_mask[ci] = fv;
if(!fv) /* not all full? enter calls bit by bit */
for(ci = 0; ci < ip->nr_k_calls; ci++)
SET_BIT(priv(rp)->s_k_call_mask,
ip->k_calls[ci]-KERNEL_CALL);
for (j = 0; j < NR_SYS_PROCS && j < BITCHUNK_BITS; j++)
if (ip->ipc_to & (1 << j))
set_sendto_bit(rp, j); /* restrict targets */
if (iskerneln(proc_nr(rp))) { /* part of the kernel? */
if (ip->stksize > 0) { /* HARDWARE stack size is 0 */
rp->p_priv->s_stack_guard = (reg_t *) ktsb;
*rp->p_priv->s_stack_guard = STACK_GUARD;
}
ktsb += ip->stksize; /* point to high end of stack */
rp->p_reg.sp = ktsb; /* this task's initial stack ptr */
hdrindex = 0; /* all use the first a.out header */
} else {
hdrindex = 1 + i-NR_TASKS; /* servers, drivers, INIT */
}
/* Architecture-specific way to find out aout header of this
* boot process.
*/
arch_get_aout_headers(hdrindex, &e_hdr);
/* Convert addresses to clicks and build process memory map */
text_base = e_hdr.a_syms >> CLICK_SHIFT;
text_clicks = (e_hdr.a_text + CLICK_SIZE-1) >> CLICK_SHIFT;
data_clicks = (e_hdr.a_data+e_hdr.a_bss + CLICK_SIZE-1) >> CLICK_SHIFT;
st_clicks= (e_hdr.a_total + CLICK_SIZE-1) >> CLICK_SHIFT;
if (!(e_hdr.a_flags & A_SEP))
{
data_clicks= (e_hdr.a_text+e_hdr.a_data+e_hdr.a_bss +
CLICK_SIZE-1) >> CLICK_SHIFT;
text_clicks = 0; /* common I&D */
}
rp->p_memmap[T].mem_phys = text_base;
rp->p_memmap[T].mem_len = text_clicks;
rp->p_memmap[D].mem_phys = text_base + text_clicks;
rp->p_memmap[D].mem_len = data_clicks;
rp->p_memmap[S].mem_phys = text_base + text_clicks + st_clicks;
rp->p_memmap[S].mem_vir = st_clicks;
rp->p_memmap[S].mem_len = 0;
/* Set initial register values. The processor status word for tasks
* is different from that of other processes because tasks can
* access I/O; this is not allowed to less-privileged processes
*/
rp->p_reg.pc = (reg_t) ip->initial_pc;
rp->p_reg.psw = (iskernelp(rp)) ? INIT_TASK_PSW : INIT_PSW;
/* Initialize the server stack pointer. Take it down one word
* to give crtso.s something to use as "argc".
*/
if (isusern(proc_nr(rp))) { /* user-space process? */
rp->p_reg.sp = (rp->p_memmap[S].mem_vir +
rp->p_memmap[S].mem_len) << CLICK_SHIFT;
rp->p_reg.sp -= sizeof(reg_t);
}
/* scheduling functions depend on proc_ptr pointing somewhere. */
if(!proc_ptr) proc_ptr = rp;
/* If this process has its own page table, VM will set the
* PT up and manage it. VM will signal the kernel when it has
* done this; until then, don't let it run.
*/
if(priv(rp)->s_flags & PROC_FULLVM)
RTS_SET(rp, VMINHIBIT);
/* Set ready. The HARDWARE task is never ready. */
if (rp->p_nr == HARDWARE) RTS_SET(rp, PROC_STOP);
RTS_UNSET(rp, SLOT_FREE); /* remove SLOT_FREE and schedule */
alloc_segments(rp);
}
/*==========================================================================*
* do_trace *
*==========================================================================*/
PUBLIC int do_trace(struct proc * caller, message * m_ptr)
{
/* Handle the debugging commands supported by the ptrace system call
* The commands are:
* T_STOP stop the process
* T_OK enable tracing by parent for this process
* T_GETINS return value from instruction space
* T_GETDATA return value from data space
* T_GETUSER return value from user process table
* T_SETINS set value in instruction space
* T_SETDATA set value in data space
* T_SETUSER set value in user process table
* T_RESUME resume execution
* T_EXIT exit
* T_STEP set trace bit
* T_SYSCALL trace system call
* T_ATTACH attach to an existing process
* T_DETACH detach from a traced process
* T_SETOPT set trace options
* T_GETRANGE get range of values
* T_SETRANGE set range of values
*
* The T_OK, T_ATTACH, T_EXIT, and T_SETOPT commands are handled completely by
* the process manager. T_GETRANGE and T_SETRANGE use sys_vircopy(). All others
* come here.
*/
register struct proc *rp;
vir_bytes tr_addr = (vir_bytes) m_ptr->CTL_ADDRESS;
long tr_data = m_ptr->CTL_DATA;
int tr_request = m_ptr->CTL_REQUEST;
int tr_proc_nr_e = m_ptr->CTL_ENDPT, tr_proc_nr;
unsigned char ub;
int i;
#define COPYTOPROC(seg, addr, myaddr, length) { \
struct vir_addr fromaddr, toaddr; \
int r; \
fromaddr.proc_nr_e = KERNEL; \
toaddr.proc_nr_e = tr_proc_nr_e; \
fromaddr.offset = (myaddr); \
toaddr.offset = (addr); \
fromaddr.segment = D; \
toaddr.segment = (seg); \
if((r=virtual_copy_vmcheck(caller, &fromaddr, \
&toaddr, length)) != OK) { \
printf("Can't copy in sys_trace: %d\n", r);\
return r;\
} \
}
#define COPYFROMPROC(seg, addr, myaddr, length) { \
struct vir_addr fromaddr, toaddr; \
int r; \
fromaddr.proc_nr_e = tr_proc_nr_e; \
toaddr.proc_nr_e = KERNEL; \
fromaddr.offset = (addr); \
toaddr.offset = (myaddr); \
fromaddr.segment = (seg); \
toaddr.segment = D; \
if((r=virtual_copy_vmcheck(caller, &fromaddr, \
&toaddr, length)) != OK) { \
printf("Can't copy in sys_trace: %d\n", r);\
return r;\
} \
}
if(!isokendpt(tr_proc_nr_e, &tr_proc_nr)) return(EINVAL);
if (iskerneln(tr_proc_nr)) return(EPERM);
rp = proc_addr(tr_proc_nr);
if (isemptyp(rp)) return(EINVAL);
switch (tr_request) {
case T_STOP: /* stop process */
RTS_SET(rp, RTS_P_STOP);
rp->p_reg.psw &= ~TRACEBIT; /* clear trace bit */
rp->p_misc_flags &= ~MF_SC_TRACE; /* clear syscall trace flag */
return(OK);
case T_GETINS: /* return value from instruction space */
COPYFROMPROC(T, tr_addr, (vir_bytes) &tr_data, sizeof(long));
m_ptr->CTL_DATA = tr_data;
break;
case T_GETDATA: /* return value from data space */
COPYFROMPROC(D, tr_addr, (vir_bytes) &tr_data, sizeof(long));
m_ptr->CTL_DATA= tr_data;
break;
case T_GETUSER: /* return value from process table */
if ((tr_addr & (sizeof(long) - 1)) != 0) return(EFAULT);
if (tr_addr <= sizeof(struct proc) - sizeof(long)) {
m_ptr->CTL_DATA = *(long *) ((char *) rp + (int) tr_addr);
break;
}
/* The process's proc struct is followed by its priv struct.
* The alignment here should be unnecessary, but better safe..
*/
i = sizeof(long) - 1;
tr_addr -= (sizeof(struct proc) + i) & ~i;
if (tr_addr > sizeof(struct priv) - sizeof(long)) return(EFAULT);
m_ptr->CTL_DATA = *(long *) ((char *) rp->p_priv + (int) tr_addr);
break;
case T_SETINS: /* set value in instruction space */
COPYTOPROC(T, tr_addr, (vir_bytes) &tr_data, sizeof(long));
m_ptr->CTL_DATA = 0;
break;
case T_SETDATA: /* set value in data space */
COPYTOPROC(D, tr_addr, (vir_bytes) &tr_data, sizeof(long));
m_ptr->CTL_DATA = 0;
break;
case T_SETUSER: /* set value in process table */
if ((tr_addr & (sizeof(reg_t) - 1)) != 0 ||
tr_addr > sizeof(struct stackframe_s) - sizeof(reg_t))
return(EFAULT);
i = (int) tr_addr;
#if (_MINIX_CHIP == _CHIP_INTEL)
/* Altering segment registers might crash the kernel when it
* tries to load them prior to restarting a process, so do
* not allow it.
*/
if (i == (int) &((struct proc *) 0)->p_reg.cs ||
i == (int) &((struct proc *) 0)->p_reg.ds ||
i == (int) &((struct proc *) 0)->p_reg.es ||
#if _WORD_SIZE == 4
i == (int) &((struct proc *) 0)->p_reg.gs ||
i == (int) &((struct proc *) 0)->p_reg.fs ||
#endif
i == (int) &((struct proc *) 0)->p_reg.ss)
return(EFAULT);
#endif
if (i == (int) &((struct proc *) 0)->p_reg.psw)
/* only selected bits are changeable */
SETPSW(rp, tr_data);
else
*(reg_t *) ((char *) &rp->p_reg + i) = (reg_t) tr_data;
m_ptr->CTL_DATA = 0;
break;
case T_DETACH: /* detach tracer */
rp->p_misc_flags &= ~MF_SC_ACTIVE;
/* fall through */
case T_RESUME: /* resume execution */
RTS_UNSET(rp, RTS_P_STOP);
m_ptr->CTL_DATA = 0;
break;
case T_STEP: /* set trace bit */
rp->p_reg.psw |= TRACEBIT;
RTS_UNSET(rp, RTS_P_STOP);
m_ptr->CTL_DATA = 0;
break;
case T_SYSCALL: /* trace system call */
rp->p_misc_flags |= MF_SC_TRACE;
RTS_UNSET(rp, RTS_P_STOP);
m_ptr->CTL_DATA = 0;
break;
case T_READB_INS: /* get value from instruction space */
COPYFROMPROC(T, tr_addr, (vir_bytes) &ub, 1);
m_ptr->CTL_DATA = ub;
break;
case T_WRITEB_INS: /* set value in instruction space */
ub = (unsigned char) (tr_data & 0xff);
COPYTOPROC(T, tr_addr, (vir_bytes) &ub, 1);
m_ptr->CTL_DATA = 0;
break;
default:
return(EINVAL);
}
return(OK);
}
