/*===========================================================================*
* swap_memreq *
*===========================================================================*/
static void swap_memreq(struct proc *src_rp, struct proc *dst_rp)
{
/* If either the source or the destination process is part of the VM request
* chain, but not both, then swap the process pointers in the chain.
*/
struct proc **rpp;
if (RTS_ISSET(src_rp, RTS_VMREQUEST) == RTS_ISSET(dst_rp, RTS_VMREQUEST))
return; /* nothing to do */
for (rpp = &vmrequest; *rpp != NULL;
rpp = &(*rpp)->p_vmrequest.nextrequestor) {
if (*rpp == src_rp) {
dst_rp->p_vmrequest.nextrequestor =
src_rp->p_vmrequest.nextrequestor;
*rpp = dst_rp;
break;
} else if (*rpp == dst_rp) {
src_rp->p_vmrequest.nextrequestor =
dst_rp->p_vmrequest.nextrequestor;
*rpp = src_rp;
break;
}
}
}
/*===========================================================================*
* 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);
}
static void kernel_call_finish(struct proc * caller, message *msg, int result)
{
if(result == VMSUSPEND) {
/* Special case: message has to be saved for handling
* until VM tells us it's allowed. VM has been notified
* and we must wait for its reply to restart the call.
*/
assert(RTS_ISSET(caller, RTS_VMREQUEST));
assert(caller->p_vmrequest.type == VMSTYPE_KERNELCALL);
caller->p_vmrequest.saved.reqmsg = *msg;
caller->p_misc_flags |= MF_KCALL_RESUME;
} else {
/*
* call is finished, we could have been suspended because of VM,
* remove the request message
*/
caller->p_vmrequest.saved.reqmsg.m_source = NONE;
if (result != EDONTREPLY) {
/* copy the result as a message to the original user buffer */
msg->m_source = SYSTEM;
msg->m_type = result; /* report status of call */
#if DEBUG_IPC_HOOK
hook_ipc_msgkresult(msg, caller);
#endif
if (copy_msg_to_user(msg, (message *)caller->p_delivermsg_vir)) {
printf("WARNING wrong user pointer 0x%08x from "
"process %s / %d\n",
caller->p_delivermsg_vir,
caller->p_name,
caller->p_endpoint);
cause_sig(proc_nr(caller), SIGSEGV);
}
}
}
}
PUBLIC void save_fpu(struct proc *pr)
{
#ifdef CONFIG_SMP
if (cpuid == pr->p_cpu) {
if (get_cpulocal_var(fpu_owner) == pr) {
disable_fpu_exception();
save_local_fpu(pr);
}
}
else {
int stopped;
/* remember if the process was already stopped */
stopped = RTS_ISSET(pr, RTS_PROC_STOP);
/* stop the remote process and force it's context to be saved */
smp_schedule_stop_proc_save_ctx(pr);
/*
* If the process wasn't stopped let the process run again. The
* process is kept block by the fact that the kernel cannot run
* on its cpu
*/
if (!stopped)
RTS_UNSET(pr, RTS_PROC_STOP);
}
#else
if (get_cpulocal_var(fpu_owner) == pr) {
disable_fpu_exception();
save_local_fpu(pr);
}
#endif
}
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));
}
void smp_schedule_stop_proc_save_ctx(struct proc * p)
{
/*
* stop the processes and force the complete context of the process to
* be saved (i.e. including FPU state and such)
*/
smp_schedule_sync(p, SCHED_IPI_STOP_PROC | SCHED_IPI_SAVE_CTX);
assert(RTS_ISSET(p, RTS_PROC_STOP));
}
void smp_schedule_migrate_proc(struct proc * p, unsigned dest_cpu)
{
/*
* stop the processes and force the complete context of the process to
* be saved (i.e. including FPU state and such)
*/
smp_schedule_sync(p, SCHED_IPI_STOP_PROC | SCHED_IPI_SAVE_CTX);
assert(RTS_ISSET(p, RTS_PROC_STOP));
/* assign the new cpu and let the process run again */
p->p_cpu = dest_cpu;
RTS_UNSET(p, RTS_PROC_STOP);
}
/*===========================================================================*
* do_endksig *
*===========================================================================*/
int do_endksig(struct proc * caller, message * m_ptr)
{
/* Finish up after a kernel type signal, caused by a SYS_KILL message or a
* call to cause_sig by a task. This is called by a signal manager after
* processing a signal it got with SYS_GETKSIG.
*/
register struct proc *rp;
int proc_nr;
/* Get process pointer and verify that it had signals pending. If the
* process is already dead its flags will be reset.
*/
if(!isokendpt(m_ptr->m_sigcalls.endpt, &proc_nr))
return EINVAL;
rp = proc_addr(proc_nr);
if (caller->p_endpoint != priv(rp)->s_sig_mgr) return(EPERM);
if (!RTS_ISSET(rp, RTS_SIG_PENDING)) return(EINVAL);
/* The signal manager has finished one kernel signal. Is the process ready? */
if (!RTS_ISSET(rp, RTS_SIGNALED)) /* new signal arrived */
RTS_UNSET(rp, RTS_SIG_PENDING); /* remove pending flag */
return(OK);
}
/*===========================================================================*
* do_setgrant *
*===========================================================================*/
int do_setgrant(struct proc * caller, message * m_ptr)
{
int r;
/* Copy grant table set in priv. struct. */
if (RTS_ISSET(caller, RTS_NO_PRIV) || !(priv(caller))) {
r = EPERM;
} else {
_K_SET_GRANT_TABLE(caller,
m_ptr->m_lsys_krn_sys_setgrant.addr,
m_ptr->m_lsys_krn_sys_setgrant.size);
r = OK;
}
return r;
}
void pagefault(struct proc *pr, int trap_errno)
{
int s;
vir_bytes ph;
u32_t pte;
if(pagefault_count != 1)
minix_panic("recursive pagefault", pagefault_count);
/* Don't schedule this process until pagefault is handled. */
if(RTS_ISSET(pr, PAGEFAULT))
minix_panic("PAGEFAULT set", pr->p_endpoint);
RTS_LOCK_SET(pr, PAGEFAULT);
if(pr->p_endpoint <= INIT_PROC_NR) {
/* Page fault we can't / don't want to
* handle.
*/
kprintf("pagefault for process %d ('%s'), pc = 0x%x\n",
pr->p_endpoint, pr->p_name, pr->p_reg.pc);
proc_stacktrace(pr);
minix_panic("page fault in system process", pr->p_endpoint);
return;
}
/* Save pagefault details, suspend process,
* add process to pagefault chain,
* and tell VM there is a pagefault to be
* handled.
*/
pr->p_pagefault.pf_virtual = pagefault_cr2;
pr->p_pagefault.pf_flags = trap_errno;
pr->p_nextpagefault = pagefaults;
pagefaults = pr;
lock_notify(HARDWARE, VM_PROC_NR);
pagefault_count = 0;
#if 0
kprintf("pagefault for process %d ('%s'), pc = 0x%x\n",
pr->p_endpoint, pr->p_name, pr->p_reg.pc);
proc_stacktrace(pr);
#endif
return;
}
/*===========================================================================*
* abort_proc_ipc_send *
*===========================================================================*/
void abort_proc_ipc_send(struct proc *rp)
{
if(RTS_ISSET(rp, RTS_SENDING)) {
struct proc **xpp;
RTS_UNSET(rp, RTS_SENDING);
rp->p_misc_flags &= ~MF_SENDING_FROM_KERNEL;
xpp = &(proc_addr(_ENDPOINT_P(rp->p_sendto_e))->p_caller_q);
while (*xpp) {
if(*xpp == rp) {
*xpp = rp->p_q_link;
rp->p_q_link = NULL;
break;
}
xpp = &(*xpp)->p_q_link;
}
}
}
/*===========================================================================*
* do_sigsend *
*===========================================================================*/
PUBLIC int do_sigsend(struct proc * caller, message * m_ptr)
{
/* Handle sys_sigsend, POSIX-style signal handling. */
struct sigmsg smsg;
register struct proc *rp;
struct sigcontext sc, *scp;
struct sigframe fr, *frp;
int proc_nr, r;
if (!isokendpt(m_ptr->SIG_ENDPT, &proc_nr)) return(EINVAL);
if (iskerneln(proc_nr)) return(EPERM);
rp = proc_addr(proc_nr);
/* Get the sigmsg structure into our address space. */
if((r=data_copy_vmcheck(caller, caller->p_endpoint,
(vir_bytes) m_ptr->SIG_CTXT_PTR, KERNEL, (vir_bytes) &smsg,
(phys_bytes) sizeof(struct sigmsg))) != OK)
return r;
/* Compute the user stack pointer where sigcontext will be stored. */
scp = (struct sigcontext *) smsg.sm_stkptr - 1;
/* Copy the registers to the sigcontext structure. */
memcpy(&sc.sc_regs, (char *) &rp->p_reg, sizeof(sigregs));
#if (_MINIX_CHIP == _CHIP_INTEL)
if(proc_used_fpu(rp)) {
/* save the FPU context before saving it to the sig context */
save_fpu(rp);
memcpy(&sc.sc_fpu_state, rp->p_fpu_state.fpu_save_area_p,
FPU_XFP_SIZE);
}
#endif
/* Finish the sigcontext initialization. */
sc.sc_mask = smsg.sm_mask;
sc.sc_flags = rp->p_misc_flags & MF_FPU_INITIALIZED;
/* Copy the sigcontext structure to the user's stack. */
if((r=data_copy_vmcheck(caller, KERNEL, (vir_bytes) &sc, m_ptr->SIG_ENDPT,
(vir_bytes) scp, (vir_bytes) sizeof(struct sigcontext))) != OK)
return r;
/* Initialize the sigframe structure. */
frp = (struct sigframe *) scp - 1;
fr.sf_scpcopy = scp;
fr.sf_retadr2= (void (*)()) rp->p_reg.pc;
fr.sf_fp = rp->p_reg.fp;
rp->p_reg.fp = (reg_t) &frp->sf_fp;
fr.sf_scp = scp;
fpu_sigcontext(rp, &fr, &sc);
fr.sf_signo = smsg.sm_signo;
fr.sf_retadr = (void (*)()) smsg.sm_sigreturn;
/* Copy the sigframe structure to the user's stack. */
if((r=data_copy_vmcheck(caller, KERNEL, (vir_bytes) &fr,
m_ptr->SIG_ENDPT, (vir_bytes) frp,
(vir_bytes) sizeof(struct sigframe))) != OK)
return r;
/* Reset user registers to execute the signal handler. */
rp->p_reg.sp = (reg_t) frp;
rp->p_reg.pc = (reg_t) smsg.sm_sighandler;
/* Signal handler should get clean FPU. */
rp->p_misc_flags &= ~MF_FPU_INITIALIZED;
if(!RTS_ISSET(rp, RTS_PROC_STOP)) {
printf("system: warning: sigsend a running process\n");
printf("caller stack: ");
proc_stacktrace(caller);
}
return(OK);
}
/*===========================================================================*
* 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;
}
}
/*
* Return the LWP status of a process, along with additional information in
* case the process is sleeping (LSSLEEP): a wchan value and text to indicate
* what the process is sleeping on, and possibly a flag field modification to
* indicate that the sleep is interruptible.
*/
static int
get_lwp_stat(int mslot, uint64_t * wcptr, char * wmptr, size_t wmsz,
int32_t * flag)
{
struct mproc *mp;
struct fproc_light *fp;
struct proc *kp;
const char *wmesg;
uint64_t wchan;
endpoint_t endpt;
mp = &mproc_tab[mslot];
fp = &fproc_tab[mslot];
kp = &proc_tab[NR_TASKS + mslot];
/*
* First cover all the cases that the process is not sleeping. In
* those cases, we need not return additional sleep information either.
*/
if (mp->mp_flags & (TRACE_ZOMBIE | ZOMBIE))
return LSZOMB;
if (mp->mp_flags & EXITING)
return LSDEAD;
if ((mp->mp_flags & TRACE_STOPPED) || RTS_ISSET(kp, RTS_P_STOP))
return LSSTOP;
if (proc_is_runnable(kp))
return LSRUN;
/*
* The process is sleeping. In that case, we must also figure out why,
* and return an appropriate wchan value and human-readable wmesg text.
*
* The process can be blocked on either a known sleep state in PM or
* VFS, or otherwise on IPC communication with another process, or
* otherwise on a kernel RTS flag. In each case, decide what to use as
* wchan value and wmesg text, and whether the sleep is interruptible.
*
* The wchan value should be unique for the sleep reason. We use its
* lower eight bits to indicate a class:
* 0x00 = kernel task
* 0x01 = kerel RTS block
* 0x02 = PM call
* 0x03 = VFS call
* 0x04 = MIB call
* 0xff = blocked on process
* The upper bits are used for class-specific information. The actual
* value does not really matter, as long as it is nonzero and there is
* no overlap between the different values.
*/
wchan = 0;
wmesg = NULL;
/*
* First see if the process is marked as blocked in the tables of PM or
* VFS. Such a block reason is always an interruptible sleep. Note
* that we do not use the kernel table at all in this case: each of the
* three tables is consistent within itself, but not necessarily
* consistent with any of the other tables, so we avoid internal
* mismatches if we can.
*/
if (mp->mp_flags & WAITING) {
wchan = 0x102;
wmesg = "wait";
} else if (mp->mp_flags & SIGSUSPENDED) {
wchan = 0x202;
wmesg = "pause";
} else if (fp->fpl_blocked_on != FP_BLOCKED_ON_NONE) {
wchan = (fp->fpl_blocked_on << 8) | 0x03;
switch (fp->fpl_blocked_on) {
case FP_BLOCKED_ON_PIPE:
wmesg = "pipe";
break;
case FP_BLOCKED_ON_FLOCK:
wmesg = "flock";
break;
case FP_BLOCKED_ON_POPEN:
wmesg = "popen";
break;
case FP_BLOCKED_ON_SELECT:
wmesg = "select";
break;
case FP_BLOCKED_ON_CDEV:
case FP_BLOCKED_ON_SDEV:
/*
* Add the task (= character or socket driver) endpoint
* to the wchan value, and use the driver's process
* name, without parentheses, as wmesg text.
*/
wchan |= (uint64_t)fp->fpl_task << 16;
fill_wmesg(wmptr, wmsz, fp->fpl_task, FALSE /*ipc*/);
break;
default:
/* A newly added flag we don't yet know about? */
wmesg = "???";
break;
}
}
if (wchan != 0) {
*wcptr = wchan;
if (wmesg != NULL) /* NULL means "already set" here */
strlcpy(wmptr, wmesg, wmsz);
*flag |= L_SINTR;
}
/*
* See if the process is blocked on sending or receiving. If not, then
* use one of the kernel RTS flags as reason.
*/
endpt = P_BLOCKEDON(kp);
switch (endpt) {
case MIB_PROC_NR:
/* This is really just aesthetics. */
wchan = 0x04;
wmesg = "sysctl";
break;
case NONE:
/*
* The process is not running, but also not blocked on IPC with
* another process. This means it must be stopped on a kernel
* RTS flag.
*/
wchan = ((uint64_t)kp->p_rts_flags << 8) | 0x01;
if (RTS_ISSET(kp, RTS_PROC_STOP))
wmesg = "kstop";
else if (RTS_ISSET(kp, RTS_SIGNALED) ||
RTS_ISSET(kp, RTS_SIGNALED))
wmesg = "ksignal";
else if (RTS_ISSET(kp, RTS_NO_PRIV))
wmesg = "knopriv";
else if (RTS_ISSET(kp, RTS_PAGEFAULT) ||
RTS_ISSET(kp, RTS_VMREQTARGET))
wmesg = "fault";
else if (RTS_ISSET(kp, RTS_NO_QUANTUM))
wmesg = "sched";
else
wmesg = "kflag";
break;
case ANY:
/*
* If the process is blocked receiving from ANY, mark it as
* being in an interruptible sleep. This looks nicer, even
* though "interruptible" is not applicable to services at all.
*/
*flag |= L_SINTR;
break;
}
/*
* If at this point wchan is still zero, the process is blocked sending
* or receiving. Use a wchan value based on the target endpoint, and
* use "(procname)" as wmesg text.
*/
if (wchan == 0) {
*wcptr = ((uint64_t)endpt << 8) | 0xff;
fill_wmesg(wmptr, wmsz, endpt, TRUE /*ipc*/);
} else {
*wcptr = wchan;
if (wmesg != NULL) /* NULL means "already set" here */
strlcpy(wmptr, wmesg, wmsz);
}
return LSSLEEP;
}
int runqueues_ok_cpu(unsigned cpu)
{
int q, l = 0;
register struct proc *xp;
struct proc **rdy_head, **rdy_tail;
rdy_head = get_cpu_var(cpu, run_q_head);
rdy_tail = get_cpu_var(cpu, run_q_tail);
for (xp = BEG_PROC_ADDR; xp < END_PROC_ADDR; ++xp) {
xp->p_found = 0;
if (l++ > MAX_LOOP) panic("check error");
}
for (q=l=0; q < NR_SCHED_QUEUES; q++) {
if (rdy_head[q] && !rdy_tail[q]) {
printf("head but no tail in %d\n", q);
return 0;
}
if (!rdy_head[q] && rdy_tail[q]) {
printf("tail but no head in %d\n", q);
return 0;
}
if (rdy_tail[q] && rdy_tail[q]->p_nextready) {
printf("tail and tail->next not null in %d\n", q);
return 0;
}
for(xp = rdy_head[q]; xp; xp = xp->p_nextready) {
const vir_bytes vxp = (vir_bytes) xp;
vir_bytes dxp;
if(vxp < (vir_bytes) BEG_PROC_ADDR || vxp >= (vir_bytes) END_PROC_ADDR) {
printf("xp out of range\n");
return 0;
}
dxp = vxp - (vir_bytes) BEG_PROC_ADDR;
if(dxp % sizeof(struct proc)) {
printf("xp not a real pointer");
return 0;
}
if(!proc_ptr_ok(xp)) {
printf("xp bogus pointer");
return 0;
}
if (RTS_ISSET(xp, RTS_SLOT_FREE)) {
printf("scheduling error: dead proc q %d %d\n",
q, xp->p_endpoint);
return 0;
}
if (!proc_is_runnable(xp)) {
printf("scheduling error: unready on runq %d proc %d\n",
q, xp->p_nr);
return 0;
}
if (xp->p_priority != q) {
printf("scheduling error: wrong priority q %d proc %d ep %d name %s\n",
q, xp->p_nr, xp->p_endpoint, xp->p_name);
return 0;
}
if (xp->p_found) {
printf("scheduling error: double sched q %d proc %d\n",
q, xp->p_nr);
return 0;
}
xp->p_found = 1;
if (!xp->p_nextready && rdy_tail[q] != xp) {
printf("sched err: last element not tail q %d proc %d\n",
q, xp->p_nr);
return 0;
}
if (l++ > MAX_LOOP) {
printf("loop in schedule queue?");
return 0;
}
}
}
for (xp = BEG_PROC_ADDR; xp < END_PROC_ADDR; ++xp) {
if(!proc_ptr_ok(xp)) {
printf("xp bogus pointer in proc table\n");
return 0;
}
if (isemptyp(xp))
continue;
if(proc_is_runnable(xp) && !xp->p_found) {
printf("sched error: ready proc %d not on queue\n", xp->p_nr);
return 0;
}
}
/* All is ok. */
return 1;
}
/*===========================================================================*
* lin_lin_copy *
*===========================================================================*/
static int lin_lin_copy(struct proc *srcproc, vir_bytes srclinaddr,
struct proc *dstproc, vir_bytes dstlinaddr, vir_bytes bytes)
{
u32_t addr;
proc_nr_t procslot;
assert(vm_running);
assert(nfreepdes >= MAX_FREEPDES);
assert(get_cpulocal_var(ptproc));
assert(get_cpulocal_var(proc_ptr));
assert(read_cr3() == get_cpulocal_var(ptproc)->p_seg.p_cr3);
procslot = get_cpulocal_var(ptproc)->p_nr;
assert(procslot >= 0 && procslot < I386_VM_DIR_ENTRIES);
if(srcproc) assert(!RTS_ISSET(srcproc, RTS_SLOT_FREE));
if(dstproc) assert(!RTS_ISSET(dstproc, RTS_SLOT_FREE));
assert(!RTS_ISSET(get_cpulocal_var(ptproc), RTS_SLOT_FREE));
assert(get_cpulocal_var(ptproc)->p_seg.p_cr3_v);
if(srcproc) assert(!RTS_ISSET(srcproc, RTS_VMINHIBIT));
if(dstproc) assert(!RTS_ISSET(dstproc, RTS_VMINHIBIT));
while(bytes > 0) {
phys_bytes srcptr, dstptr;
vir_bytes chunk = bytes;
int changed = 0;
#ifdef CONFIG_SMP
unsigned cpu = cpuid;
if (GET_BIT(srcproc->p_stale_tlb, cpu)) {
changed = 1;
UNSET_BIT(srcproc->p_stale_tlb, cpu);
}
if (GET_BIT(dstproc->p_stale_tlb, cpu)) {
changed = 1;
UNSET_BIT(dstproc->p_stale_tlb, cpu);
}
#endif
/* Set up 4MB ranges. */
srcptr = createpde(srcproc, srclinaddr, &chunk, 0, &changed);
dstptr = createpde(dstproc, dstlinaddr, &chunk, 1, &changed);
if(changed)
reload_cr3();
/* Copy pages. */
PHYS_COPY_CATCH(srcptr, dstptr, chunk, addr);
if(addr) {
/* If addr is nonzero, a page fault was caught. */
if(addr >= srcptr && addr < (srcptr + chunk)) {
return EFAULT_SRC;
}
if(addr >= dstptr && addr < (dstptr + chunk)) {
return EFAULT_DST;
}
panic("lin_lin_copy fault out of range");
/* Not reached. */
return EFAULT;
}
/* Update counter and addresses for next iteration, if any. */
bytes -= chunk;
srclinaddr += chunk;
dstlinaddr += chunk;
}
if(srcproc) assert(!RTS_ISSET(srcproc, RTS_SLOT_FREE));
if(dstproc) assert(!RTS_ISSET(dstproc, RTS_SLOT_FREE));
assert(!RTS_ISSET(get_cpulocal_var(ptproc), RTS_SLOT_FREE));
assert(get_cpulocal_var(ptproc)->p_seg.p_cr3_v);
return OK;
}
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;
}
/*===========================================================================*
* lin_lin_copy *
*===========================================================================*/
static int lin_lin_copy(struct proc *srcproc, vir_bytes srclinaddr,
struct proc *dstproc, vir_bytes dstlinaddr, vir_bytes bytes)
{
u32_t addr;
proc_nr_t procslot;
assert(get_cpulocal_var(ptproc));
assert(get_cpulocal_var(proc_ptr));
assert(read_ttbr0() == get_cpulocal_var(ptproc)->p_seg.p_ttbr);
procslot = get_cpulocal_var(ptproc)->p_nr;
assert(procslot >= 0 && procslot < ARM_VM_DIR_ENTRIES);
if(srcproc) assert(!RTS_ISSET(srcproc, RTS_SLOT_FREE));
if(dstproc) assert(!RTS_ISSET(dstproc, RTS_SLOT_FREE));
assert(!RTS_ISSET(get_cpulocal_var(ptproc), RTS_SLOT_FREE));
assert(get_cpulocal_var(ptproc)->p_seg.p_ttbr_v);
if(srcproc) assert(!RTS_ISSET(srcproc, RTS_VMINHIBIT));
if(dstproc) assert(!RTS_ISSET(dstproc, RTS_VMINHIBIT));
while(bytes > 0) {
phys_bytes srcptr, dstptr;
vir_bytes chunk = bytes;
int changed = 0;
#ifdef CONFIG_SMP
unsigned cpu = cpuid;
if (srcproc && GET_BIT(srcproc->p_stale_tlb, cpu)) {
changed = 1;
UNSET_BIT(srcproc->p_stale_tlb, cpu);
}
if (dstproc && GET_BIT(dstproc->p_stale_tlb, cpu)) {
changed = 1;
UNSET_BIT(dstproc->p_stale_tlb, cpu);
}
#endif
/* Set up 1MB ranges. */
srcptr = createpde(srcproc, srclinaddr, &chunk, 0, &changed);
dstptr = createpde(dstproc, dstlinaddr, &chunk, 1, &changed);
if(changed) {
reload_ttbr0();
}
/* Copy pages. */
PHYS_COPY_CATCH(srcptr, dstptr, chunk, addr);
if(addr) {
/* If addr is nonzero, a page fault was caught.
*
* phys_copy does all memory accesses word-aligned (rounded
* down), so pagefaults can occur at a lower address than
* the specified offsets. compute the lower bounds for sanity
* check use.
*/
vir_bytes src_aligned = srcptr & ~0x3, dst_aligned = dstptr & ~0x3;
if(addr >= src_aligned && addr < (srcptr + chunk)) {
return EFAULT_SRC;
}
if(addr >= dst_aligned && addr < (dstptr + chunk)) {
return EFAULT_DST;
}
panic("lin_lin_copy fault out of range");
/* Not reached. */
return EFAULT;
}
/* Update counter and addresses for next iteration, if any. */
bytes -= chunk;
srclinaddr += chunk;
dstlinaddr += chunk;
}
if(srcproc) assert(!RTS_ISSET(srcproc, RTS_SLOT_FREE));
if(dstproc) assert(!RTS_ISSET(dstproc, RTS_SLOT_FREE));
assert(!RTS_ISSET(get_cpulocal_var(ptproc), RTS_SLOT_FREE));
assert(get_cpulocal_var(ptproc)->p_seg.p_ttbr_v);
return OK;
}
/*===========================================================================*
* sys_task *
*===========================================================================*/
PUBLIC void sys_task()
{
/* Main entry point of sys_task. Get the message and dispatch on type. */
static message m;
register int result;
register struct proc *caller_ptr;
int s;
int call_nr;
int n = 0;
/* Initialize the system task. */
initialize();
while (TRUE) {
struct proc *restarting;
restarting = vmrestart_check(&m);
if(!restarting) {
int r;
/* Get work. Block and wait until a request message arrives. */
if((r=receive(ANY, &m)) != OK)
minix_panic("receive() failed", r);
}
sys_call_code = (unsigned) m.m_type;
call_nr = sys_call_code - KERNEL_CALL;
who_e = m.m_source;
okendpt(who_e, &who_p);
caller_ptr = proc_addr(who_p);
/* See if the caller made a valid request and try to handle it. */
if (call_nr < 0 || call_nr >= NR_SYS_CALLS) { /* check call number */
kprintf("SYSTEM: illegal request %d from %d.\n",
call_nr,m.m_source);
result = EBADREQUEST; /* illegal message type */
}
else if (!GET_BIT(priv(caller_ptr)->s_k_call_mask, call_nr)) {
result = ECALLDENIED; /* illegal message type */
}
else {
result = (*call_vec[call_nr])(&m); /* handle the system call */
}
if(result == VMSUSPEND) {
/* Special case: message has to be saved for handling
* until VM tells us it's allowed. VM has been notified
* and we must wait for its reply to restart the call.
*/
vmassert(RTS_ISSET(caller_ptr, VMREQUEST));
vmassert(caller_ptr->p_vmrequest.type == VMSTYPE_KERNELCALL);
memcpy(&caller_ptr->p_vmrequest.saved.reqmsg, &m, sizeof(m));
} else if (result != EDONTREPLY) {
/* Send a reply, unless inhibited by a handler function.
* Use the kernel function lock_send() to prevent a system
* call trap.
*/
if(restarting) {
vmassert(!RTS_ISSET(restarting, VMREQUEST));
#if 0
vmassert(!RTS_ISSET(restarting, VMREQTARGET));
#endif
}
m.m_type = result; /* report status of call */
if(WILLRECEIVE(caller_ptr, SYSTEM)) {
if (OK != (s=lock_send(m.m_source, &m))) {
kprintf("SYSTEM, reply to %d failed: %d\n",
m.m_source, s);
}
} else {
kprintf("SYSTEM: not replying to %d; not ready\n",
caller_ptr->p_endpoint);
}
}
}
}
/*===========================================================================*
* do_sigsend *
*===========================================================================*/
int do_sigsend(struct proc * caller, message * m_ptr)
{
/* Handle sys_sigsend, POSIX-style signal handling. */
struct sigmsg smsg;
register struct proc *rp;
struct sigcontext sc, *scp;
struct sigframe fr, *frp;
int proc_nr, r;
if (!isokendpt(m_ptr->SIG_ENDPT, &proc_nr)) return(EINVAL);
if (iskerneln(proc_nr)) return(EPERM);
rp = proc_addr(proc_nr);
/* Get the sigmsg structure into our address space. */
if((r=data_copy_vmcheck(caller, caller->p_endpoint,
(vir_bytes) m_ptr->SIG_CTXT_PTR, KERNEL, (vir_bytes) &smsg,
(phys_bytes) sizeof(struct sigmsg))) != OK)
return r;
/* Compute the user stack pointer where sigcontext will be stored. */
smsg.sm_stkptr = arch_get_sp(rp);
scp = (struct sigcontext *) smsg.sm_stkptr - 1;
/* Copy the registers to the sigcontext structure. */
memcpy(&sc.sc_regs, (char *) &rp->p_reg, sizeof(sigregs));
#if defined(__i386__)
sc.trap_style = rp->p_seg.p_kern_trap_style;
if(sc.trap_style == KTS_NONE) {
printf("do_sigsend: sigsend an unsaved process\n");
return EINVAL;
}
if(proc_used_fpu(rp)) {
/* save the FPU context before saving it to the sig context */
save_fpu(rp);
memcpy(&sc.sc_fpu_state, rp->p_seg.fpu_state, FPU_XFP_SIZE);
}
#endif
/* Finish the sigcontext initialization. */
sc.sc_mask = smsg.sm_mask;
sc.sc_flags = rp->p_misc_flags & MF_FPU_INITIALIZED;
/* Copy the sigcontext structure to the user's stack. */
if((r=data_copy_vmcheck(caller, KERNEL, (vir_bytes) &sc, m_ptr->SIG_ENDPT,
(vir_bytes) scp, (vir_bytes) sizeof(struct sigcontext))) != OK)
return r;
/* Initialize the sigframe structure. */
frp = (struct sigframe *) scp - 1;
fr.sf_scpcopy = scp;
fr.sf_retadr2= (void (*)()) rp->p_reg.pc;
fr.sf_fp = rp->p_reg.fp;
rp->p_reg.fp = (reg_t) &frp->sf_fp;
fr.sf_scp = scp;
fpu_sigcontext(rp, &fr, &sc);
fr.sf_signo = smsg.sm_signo;
fr.sf_retadr = (void (*)()) smsg.sm_sigreturn;
#if defined(__arm__)
/* use the ARM link register to set the return address from the signal
* handler
*/
rp->p_reg.lr = (reg_t) fr.sf_retadr;
if(rp->p_reg.lr & 1) { printf("sigsend: LSB LR makes no sense.\n"); }
/* pass signal handler parameters in registers */
rp->p_reg.retreg = (reg_t) fr.sf_signo;
rp->p_reg.r1 = (reg_t) fr.sf_code;
rp->p_reg.r2 = (reg_t) fr.sf_scp;
rp->p_misc_flags |= MF_CONTEXT_SET;
#endif
/* Copy the sigframe structure to the user's stack. */
if((r=data_copy_vmcheck(caller, KERNEL, (vir_bytes) &fr,
m_ptr->SIG_ENDPT, (vir_bytes) frp,
(vir_bytes) sizeof(struct sigframe))) != OK)
return r;
/* Reset user registers to execute the signal handler. */
rp->p_reg.sp = (reg_t) frp;
rp->p_reg.pc = (reg_t) smsg.sm_sighandler;
/* Signal handler should get clean FPU. */
rp->p_misc_flags &= ~MF_FPU_INITIALIZED;
if(!RTS_ISSET(rp, RTS_PROC_STOP)) {
printf("system: warning: sigsend a running process\n");
printf("caller stack: ");
proc_stacktrace(caller);
}
return(OK);
}
/*===========================================================================*
* 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);
}
