/*===========================================================================*
* pm_fork *
*===========================================================================*/
void pm_fork(endpoint_t pproc, endpoint_t cproc, pid_t cpid)
{
/* Perform those aspects of the fork() system call that relate to files.
* In particular, let the child inherit its parent's file descriptors.
* The parent and child parameters tell who forked off whom. The file
* system uses the same slot numbers as the kernel. Only PM makes this call.
*/
struct fproc *cp, *pp;
int i, parentno, childno;
mutex_t c_fp_lock;
/* Check up-to-dateness of fproc. */
okendpt(pproc, &parentno);
/* PM gives child endpoint, which implies process slot information.
* Don't call isokendpt, because that will verify if the endpoint
* number is correct in fproc, which it won't be.
*/
childno = _ENDPOINT_P(cproc);
if (childno < 0 || childno >= NR_PROCS)
panic("VFS: bogus child for forking: %d", cproc);
if (fproc[childno].fp_pid != PID_FREE)
panic("VFS: forking on top of in-use child: %d", childno);
/* Copy the parent's fproc struct to the child. */
/* However, the mutex variables belong to a slot and must stay the same. */
c_fp_lock = fproc[childno].fp_lock;
fproc[childno] = fproc[parentno];
fproc[childno].fp_lock = c_fp_lock;
/* Increase the counters in the 'filp' table. */
cp = &fproc[childno];
pp = &fproc[parentno];
for (i = 0; i < OPEN_MAX; i++)
if (cp->fp_filp[i] != NULL) cp->fp_filp[i]->filp_count++;
/* Fill in new process and endpoint id. */
cp->fp_pid = cpid;
cp->fp_endpoint = cproc;
/* A forking process never has an outstanding grant, as it isn't blocking on
* I/O. */
if (GRANT_VALID(pp->fp_grant)) {
panic("VFS: fork: pp (endpoint %d) has grant %d\n", pp->fp_endpoint,
pp->fp_grant);
}
if (GRANT_VALID(cp->fp_grant)) {
panic("VFS: fork: cp (endpoint %d) has grant %d\n", cp->fp_endpoint,
cp->fp_grant);
}
/* A child is not a process leader, not being revived, etc. */
cp->fp_flags = FP_NOFLAGS;
/* Record the fact that both root and working dir have another user. */
if (cp->fp_rd) dup_vnode(cp->fp_rd);
if (cp->fp_wd) dup_vnode(cp->fp_wd);
}
/*===========================================================================*
* 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;
unsigned int call_nr;
int s;
/* Initialize the system task. */
initialize();
while (TRUE) {
/* Get work. Block and wait until a request message arrives. */
receive(ANY, &m);
call_nr = (unsigned) m.m_type - 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 (! (priv(caller_ptr)->s_call_mask & (1<<call_nr))) {
#if DEBUG_ENABLE_IPC_WARNINGS
kprintf("SYSTEM: request %d from %d denied.\n", call_nr,m.m_source);
#endif
result = ECALLDENIED; /* illegal message type */
} else if (call_nr >= NR_SYS_CALLS) { /* check call number */
#if DEBUG_ENABLE_IPC_WARNINGS
kprintf("SYSTEM: illegal request %d from %d.\n", call_nr,m.m_source);
#endif
result = EBADREQUEST; /* illegal message type */
}
else {
result = (*call_vec[call_nr])(&m); /* handle the system call */
}
/* Send a reply, unless inhibited by a handler function. Use the kernel
* function lock_send() to prevent a system call trap. The destination
* is known to be blocked waiting for a message.
*/
if (result != EDONTREPLY) {
m.m_type = result; /* report status of call */
if (OK != (s=lock_send(m.m_source, &m))) {
kprintf("SYSTEM, reply to %d failed: %d\n", m.m_source, s);
}
}
}
}
/*===========================================================================*
* 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);
}
/*===========================================================================*
* service_pm *
*===========================================================================*/
static void service_pm()
{
int r, slot;
switch (job_call_nr) {
case PM_SETUID:
{
endpoint_t proc_e;
uid_t euid, ruid;
proc_e = job_m_in.PM_PROC;
euid = job_m_in.PM_EID;
ruid = job_m_in.PM_RID;
pm_setuid(proc_e, euid, ruid);
m_out.m_type = PM_SETUID_REPLY;
m_out.PM_PROC = proc_e;
}
break;
case PM_SETGID:
{
endpoint_t proc_e;
gid_t egid, rgid;
proc_e = job_m_in.PM_PROC;
egid = job_m_in.PM_EID;
rgid = job_m_in.PM_RID;
pm_setgid(proc_e, egid, rgid);
m_out.m_type = PM_SETGID_REPLY;
m_out.PM_PROC = proc_e;
}
break;
case PM_SETSID:
{
endpoint_t proc_e;
proc_e = job_m_in.PM_PROC;
pm_setsid(proc_e);
m_out.m_type = PM_SETSID_REPLY;
m_out.PM_PROC = proc_e;
}
break;
case PM_EXEC:
case PM_EXIT:
case PM_DUMPCORE:
{
endpoint_t proc_e = job_m_in.PM_PROC;
okendpt(proc_e, &slot);
fp = &fproc[slot];
if (fp->fp_flags & FP_PENDING) {
/* This process has a request pending, but PM wants it
* gone. Forget about the pending request and satisfy
* PM's request instead. Note that a pending request
* AND an EXEC request are mutually exclusive. Also, PM
* should send only one request/process at a time.
*/
assert(fp->fp_job.j_m_in.m_source != PM_PROC_NR);
}
/* PM requests on behalf of a proc are handled after the
* system call that might be in progress for that proc has
* finished. If the proc is not busy, we start a dummy call.
*/
if (!(fp->fp_flags & FP_PENDING) &&
mutex_trylock(&fp->fp_lock) == 0) {
mutex_unlock(&fp->fp_lock);
worker_start(do_dummy);
fp->fp_flags |= FP_DROP_WORK;
}
fp->fp_job.j_m_in = job_m_in;
fp->fp_flags |= FP_PM_PENDING;
return;
}
case PM_FORK:
case PM_SRV_FORK:
{
endpoint_t pproc_e, proc_e;
pid_t child_pid;
uid_t reuid;
gid_t regid;
pproc_e = job_m_in.PM_PPROC;
proc_e = job_m_in.PM_PROC;
child_pid = job_m_in.PM_CPID;
reuid = job_m_in.PM_REUID;
regid = job_m_in.PM_REGID;
pm_fork(pproc_e, proc_e, child_pid);
m_out.m_type = PM_FORK_REPLY;
if (job_call_nr == PM_SRV_FORK) {
m_out.m_type = PM_SRV_FORK_REPLY;
pm_setuid(proc_e, reuid, reuid);
pm_setgid(proc_e, regid, regid);
}
m_out.PM_PROC = proc_e;
}
break;
case PM_SETGROUPS:
{
endpoint_t proc_e;
int group_no;
gid_t *group_addr;
proc_e = job_m_in.PM_PROC;
group_no = job_m_in.PM_GROUP_NO;
group_addr = (gid_t *) job_m_in.PM_GROUP_ADDR;
pm_setgroups(proc_e, group_no, group_addr);
m_out.m_type = PM_SETGROUPS_REPLY;
m_out.PM_PROC = proc_e;
}
break;
case PM_UNPAUSE:
{
endpoint_t proc_e;
proc_e = job_m_in.PM_PROC;
unpause(proc_e);
m_out.m_type = PM_UNPAUSE_REPLY;
m_out.PM_PROC = proc_e;
}
break;
case PM_REBOOT:
pm_reboot();
/* Reply dummy status to PM for synchronization */
m_out.m_type = PM_REBOOT_REPLY;
break;
default:
printf("VFS: don't know how to handle PM request %d\n", job_call_nr);
return;
}
r = send(PM_PROC_NR, &m_out);
if (r != OK)
panic("service_pm: send failed: %d", r);
}
/*===========================================================================*
* 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);
}
/*===========================================================================*
* 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);
}
}
}
}
/*===========================================================================*
* pm_exec *
*===========================================================================*/
int pm_exec(endpoint_t proc_e, vir_bytes path, size_t path_len,
vir_bytes frame, size_t frame_len, vir_bytes *pc,
vir_bytes *newsp, int user_exec_flags)
{
/* Perform the execve(name, argv, envp) call. The user library builds a
* complete stack image, including pointers, args, environ, etc. The stack
* is copied to a buffer inside VFS, and then to the new core image.
*/
int r, slot;
vir_bytes vsp;
struct fproc *rfp;
int extrabase = 0;
static char mbuf[ARG_MAX]; /* buffer for stack and zeroes */
struct vfs_exec_info execi;
int i;
static char fullpath[PATH_MAX],
elf_interpreter[PATH_MAX],
firstexec[PATH_MAX],
finalexec[PATH_MAX];
struct lookup resolve;
struct fproc *vmfp = &fproc[VM_PROC_NR];
stackhook_t makestack = NULL;
static int n;
n++;
struct filp *newfilp = NULL;
lock_exec();
lock_proc(vmfp, 0);
/* unset execi values are 0. */
memset(&execi, 0, sizeof(execi));
execi.vmfd = -1;
/* passed from exec() libc code */
execi.userflags = user_exec_flags;
execi.args.stack_high = kinfo.user_sp;
execi.args.stack_size = DEFAULT_STACK_LIMIT;
okendpt(proc_e, &slot);
rfp = fp = &fproc[slot];
lookup_init(&resolve, fullpath, PATH_NOFLAGS, &execi.vmp, &execi.vp);
resolve.l_vmnt_lock = VMNT_READ;
resolve.l_vnode_lock = VNODE_READ;
/* Fetch the stack from the user before destroying the old core image. */
if (frame_len > ARG_MAX)
FAILCHECK(ENOMEM); /* stack too big */
r = sys_datacopy(proc_e, (vir_bytes) frame, SELF, (vir_bytes) mbuf,
(size_t) frame_len);
if (r != OK) { /* can't fetch stack (e.g. bad virtual addr) */
printf("VFS: pm_exec: sys_datacopy failed\n");
FAILCHECK(r);
}
/* The default is to keep the original user and group IDs */
execi.args.new_uid = rfp->fp_effuid;
execi.args.new_gid = rfp->fp_effgid;
/* Get the exec file name. */
FAILCHECK(fetch_name(path, path_len, fullpath));
strlcpy(finalexec, fullpath, PATH_MAX);
strlcpy(firstexec, fullpath, PATH_MAX);
/* Get_read_vp will return an opened vn in execi.
* if necessary it releases the existing vp so we can
* switch after we find out what's inside the file.
* It reads the start of the file.
*/
Get_read_vp(execi, fullpath, 1, 1, &resolve, fp);
/* If this is a script (i.e. has a #!/interpreter line),
* retrieve the name of the interpreter and open that
* executable instead.
*/
if(is_script(&execi)) {
/* patch_stack will add interpreter name and
* args to stack and retrieve the new binary
* name into fullpath.
*/
FAILCHECK(fetch_name(path, path_len, fullpath));
FAILCHECK(patch_stack(execi.vp, mbuf, &frame_len, fullpath));
strlcpy(finalexec, fullpath, PATH_MAX);
strlcpy(firstexec, fullpath, PATH_MAX);
Get_read_vp(execi, fullpath, 1, 0, &resolve, fp);
}
/* If this is a dynamically linked executable, retrieve
* the name of that interpreter in elf_interpreter and open that
* executable instead. But open the current executable in an
* fd for the current process.
*/
if(elf_has_interpreter(execi.args.hdr, execi.args.hdr_len,
elf_interpreter, sizeof(elf_interpreter))) {
/* Switch the executable vnode to the interpreter */
execi.is_dyn = 1;
/* The interpreter (loader) needs an fd to the main program,
* which is currently in finalexec
*/
if((r = execi.elf_main_fd = common_open(finalexec, O_RDONLY, 0)) < 0) {
printf("VFS: exec: dynamic: open main exec failed %s (%d)\n",
fullpath, r);
FAILCHECK(r);
}
/* ld.so is linked at 0, but it can relocate itself; we
* want it higher to trap NULL pointer dereferences.
*/
execi.args.load_offset = 0x10000;
/* Remember it */
strlcpy(execi.execname, finalexec, PATH_MAX);
/* The executable we need to execute first (loader)
* is in elf_interpreter, and has to be in fullpath to
* be looked up
*/
strlcpy(fullpath, elf_interpreter, PATH_MAX);
strlcpy(firstexec, elf_interpreter, PATH_MAX);
Get_read_vp(execi, fullpath, 0, 0, &resolve, fp);
}
/* We also want an FD for VM to mmap() the process in if possible. */
{
struct vnode *vp = execi.vp;
assert(vp);
if(vp->v_vmnt->m_haspeek && major(vp->v_dev) != MEMORY_MAJOR) {
int newfd = -1;
if(get_fd(vmfp, 0, R_BIT, &newfd, &newfilp) == OK) {
assert(newfd >= 0 && newfd < OPEN_MAX);
assert(!vmfp->fp_filp[newfd]);
newfilp->filp_count = 1;
newfilp->filp_vno = vp;
newfilp->filp_flags = O_RDONLY;
FD_SET(newfd, &vmfp->fp_filp_inuse);
vmfp->fp_filp[newfd] = newfilp;
/* dup_vnode(vp); */
execi.vmfd = newfd;
execi.args.memmap = vfs_memmap;
}
}
}
/* callback functions and data */
execi.args.copymem = read_seg;
execi.args.clearproc = libexec_clearproc_vm_procctl;
execi.args.clearmem = libexec_clear_sys_memset;
execi.args.allocmem_prealloc_cleared = libexec_alloc_mmap_prealloc_cleared;
execi.args.allocmem_prealloc_junk = libexec_alloc_mmap_prealloc_junk;
execi.args.allocmem_ondemand = libexec_alloc_mmap_ondemand;
execi.args.opaque = &execi;
execi.args.proc_e = proc_e;
execi.args.frame_len = frame_len;
execi.args.filesize = execi.vp->v_size;
for (i = 0; exec_loaders[i].load_object != NULL; i++) {
r = (*exec_loaders[i].load_object)(&execi.args);
/* Loaded successfully, so no need to try other loaders */
if (r == OK) { makestack = exec_loaders[i].setup_stack; break; }
}
FAILCHECK(r);
/* Inform PM */
FAILCHECK(libexec_pm_newexec(proc_e, &execi.args));
/* Save off PC */
*pc = execi.args.pc;
/* call a stack-setup function if this executable type wants it */
vsp = execi.args.stack_high - frame_len;
if(makestack) FAILCHECK(makestack(&execi, mbuf, &frame_len, &vsp, &extrabase));
/* Patch up stack and copy it from VFS to new core image. */
libexec_patch_ptr(mbuf, vsp + extrabase);
FAILCHECK(sys_datacopy(SELF, (vir_bytes) mbuf, proc_e, (vir_bytes) vsp,
(phys_bytes)frame_len));
/* Return new stack pointer to caller */
*newsp = vsp;
clo_exec(rfp);
if (execi.args.allow_setuid) {
/* If after loading the image we're still allowed to run with
* setuid or setgid, change credentials now */
rfp->fp_effuid = execi.args.new_uid;
rfp->fp_effgid = execi.args.new_gid;
}
/* Remember the new name of the process */
strlcpy(rfp->fp_name, execi.args.progname, PROC_NAME_LEN);
pm_execfinal:
if(newfilp) unlock_filp(newfilp);
else if (execi.vp != NULL) {
unlock_vnode(execi.vp);
put_vnode(execi.vp);
}
if(execi.vmfd >= 0 && !execi.vmfd_used) {
if(OK != close_fd(vmfp, execi.vmfd)) {
printf("VFS: unexpected close fail of vm fd\n");
}
}
unlock_proc(vmfp);
unlock_exec();
return(r);
}
/*===========================================================================*
* do_vumap *
*===========================================================================*/
int do_vumap(struct proc *caller, message *m_ptr)
{
/* Map a vector of grants or local virtual addresses to physical addresses.
* Designed to be used by drivers to perform an efficient lookup of physical
* addresses for the purpose of direct DMA from/to a remote process.
*/
endpoint_t endpt, source, granter;
struct proc *procp;
struct vumap_vir vvec[MAPVEC_NR];
struct vumap_phys pvec[MAPVEC_NR];
vir_bytes vaddr, paddr, vir_addr;
phys_bytes phys_addr;
int i, r, proc_nr, vcount, pcount, pmax, access;
size_t size, chunk, offset;
endpt = caller->p_endpoint;
/* Retrieve and check input parameters. */
source = m_ptr->VUMAP_ENDPT;
vaddr = (vir_bytes) m_ptr->VUMAP_VADDR;
vcount = m_ptr->VUMAP_VCOUNT;
offset = m_ptr->VUMAP_OFFSET;
access = m_ptr->VUMAP_ACCESS;
paddr = (vir_bytes) m_ptr->VUMAP_PADDR;
pmax = m_ptr->VUMAP_PMAX;
if (vcount <= 0 || pmax <= 0)
return EINVAL;
if (vcount > MAPVEC_NR) vcount = MAPVEC_NR;
if (pmax > MAPVEC_NR) pmax = MAPVEC_NR;
/* Convert access to safecopy access flags. */
switch (access) {
case VUA_READ: access = CPF_READ; break;
case VUA_WRITE: access = CPF_WRITE; break;
case VUA_READ|VUA_WRITE: access = CPF_READ|CPF_WRITE; break;
default: return EINVAL;
}
/* Copy in the vector of virtual addresses. */
size = vcount * sizeof(vvec[0]);
if (data_copy(endpt, vaddr, KERNEL, (vir_bytes) vvec, size) != OK)
return EFAULT;
pcount = 0;
/* Go through the input entries, one at a time. Stop early in case the output
* vector has filled up.
*/
for (i = 0; i < vcount && pcount < pmax; i++) {
size = vvec[i].vv_size;
if (size <= offset)
return EINVAL;
size -= offset;
if (source != SELF) {
r = verify_grant(source, endpt, vvec[i].vv_grant, size, access,
offset, &vir_addr, &granter);
if (r != OK)
return r;
} else {
vir_addr = vvec[i].vv_addr + offset;
granter = endpt;
}
okendpt(granter, &proc_nr);
procp = proc_addr(proc_nr);
/* Each virtual range is made up of one or more physical ranges. */
while (size > 0 && pcount < pmax) {
chunk = vm_lookup_range(procp, vir_addr, &phys_addr, size);
if (!chunk) {
/* Try to get the memory allocated, unless the memory
* is supposed to be there to be read from.
*/
if (access & CPF_READ)
return EFAULT;
/* This call may suspend the current call, or return an
* error for a previous invocation.
*/
return vm_check_range(caller, procp, vir_addr, size);
}
pvec[pcount].vp_addr = phys_addr;
pvec[pcount].vp_size = chunk;
pcount++;
vir_addr += chunk;
size -= chunk;
}
offset = 0;
}
/* Copy out the resulting vector of physical addresses. */
assert(pcount > 0);
size = pcount * sizeof(pvec[0]);
r = data_copy_vmcheck(caller, KERNEL, (vir_bytes) pvec, endpt, paddr, size);
if (r == OK)
m_ptr->VUMAP_PCOUNT = pcount;
return r;
}
/*===========================================================================*
* pm_exec *
*===========================================================================*/
PUBLIC int pm_exec(int proc_e, char *path, vir_bytes path_len, char *frame,
vir_bytes frame_len, vir_bytes *pc)
{
/* Perform the execve(name, argv, envp) call. The user library builds a
* complete stack image, including pointers, args, environ, etc. The stack
* is copied to a buffer inside VFS, and then to the new core image.
*/
int r, r1, round, proc_s;
vir_bytes vsp;
struct fproc *rfp;
struct vnode *vp;
char *cp;
static char mbuf[ARG_MAX]; /* buffer for stack and zeroes */
struct exec_info execi;
int i;
okendpt(proc_e, &proc_s);
rfp = fp = &fproc[proc_s];
who_e = proc_e;
who_p = proc_s;
super_user = (fp->fp_effuid == SU_UID ? TRUE : FALSE); /* su? */
/* Get the exec file name. */
if ((r = fetch_name(path, path_len, 0)) != OK) return(r);
/* Fetch the stack from the user before destroying the old core image. */
if (frame_len > ARG_MAX) {
printf("VFS: pm_exec: stack too big\n");
return(ENOMEM); /* stack too big */
}
r = sys_datacopy(proc_e, (vir_bytes) frame, SELF, (vir_bytes) mbuf,
(phys_bytes) frame_len);
if (r != OK) { /* can't fetch stack (e.g. bad virtual addr) */
printf("pm_exec: sys_datacopy failed\n");
return(r);
}
/* The default is to keep the original user and group IDs */
execi.new_uid = rfp->fp_effuid;
execi.new_gid = rfp->fp_effgid;
for (round= 0; round < 2; round++) {
/* round = 0 (first attempt), or 1 (interpreted script) */
/* Save the name of the program */
(cp= strrchr(user_fullpath, '/')) ? cp++ : (cp= user_fullpath);
strncpy(execi.progname, cp, PROC_NAME_LEN-1);
execi.progname[PROC_NAME_LEN-1] = '\0';
execi.setugid = 0;
/* Open executable */
if ((vp = eat_path(PATH_NOFLAGS, fp)) == NULL) return(err_code);
execi.vp = vp;
if ((vp->v_mode & I_TYPE) != I_REGULAR)
r = ENOEXEC;
else if ((r1 = forbidden(vp, X_BIT)) != OK)
r = r1;
else
r = req_stat(vp->v_fs_e, vp->v_inode_nr, VFS_PROC_NR,
(char *) &(execi.sb), 0, 0);
if (r != OK) {
put_vnode(vp);
return(r);
}
if (round == 0) {
/* Deal with setuid/setgid executables */
if (vp->v_mode & I_SET_UID_BIT) {
execi.new_uid = vp->v_uid;
execi.setugid = 1;
}
if (vp->v_mode & I_SET_GID_BIT) {
execi.new_gid = vp->v_gid;
execi.setugid = 1;
}
}
r = map_header(&execi.hdr, execi.vp);
if (r != OK) {
put_vnode(vp);
return(r);
}
if (!is_script(execi.hdr, execi.vp->v_size) || round != 0)
break;
/* Get fresh copy of the file name. */
if ((r = fetch_name(path, path_len, 0)) != OK)
printf("VFS pm_exec: 2nd fetch_name failed\n");
else if ((r = patch_stack(vp, mbuf, &frame_len)) != OK)
printf("VFS pm_exec: patch_stack failed\n");
put_vnode(vp);
if (r != OK) return(r);
}
execi.proc_e = proc_e;
execi.frame_len = frame_len;
for(i = 0; exec_loaders[i].load_object != NULL; i++) {
r = (*exec_loaders[i].load_object)(&execi);
/* Loaded successfully, so no need to try other loaders */
if (r == OK) break;
}
put_vnode(vp);
/* No exec loader could load the object */
if (r != OK) {
return(ENOEXEC);
}
/* Save off PC */
*pc = execi.pc;
/* Patch up stack and copy it from VFS to new core image. */
vsp = execi.stack_top;
vsp -= frame_len;
patch_ptr(mbuf, vsp);
if ((r = sys_datacopy(SELF, (vir_bytes) mbuf, proc_e, (vir_bytes) vsp,
(phys_bytes)frame_len)) != OK) {
printf("VFS: datacopy failed (%d) trying to copy to %lu\n", r, vsp);
return(r);
}
if (r != OK) return(r);
clo_exec(rfp);
if (execi.setugid) {
/* If after loading the image we're still allowed to run with
* setuid or setgid, change the credentials now */
rfp->fp_effuid = execi.new_uid;
rfp->fp_effgid = execi.new_gid;
}
/* This child has now exec()ced. */
rfp->fp_execced = 1;
return(OK);
}
