/*===========================================================================*
* fs_breadwrite *
*===========================================================================*/
PUBLIC int fs_breadwrite(void)
{
int r, rw_flag, completed;
cp_grant_id_t gid;
u64_t position;
unsigned int off, cum_io, chunk, block_size;
size_t nrbytes;
/* Pseudo inode for rw_chunk */
struct inode rip;
r = OK;
/* Get the values from the request message */
rw_flag = (fs_m_in.m_type == REQ_BREAD ? READING : WRITING);
gid = (cp_grant_id_t) fs_m_in.REQ_GRANT;
position = make64((unsigned long) fs_m_in.REQ_SEEK_POS_LO,
(unsigned long) fs_m_in.REQ_SEEK_POS_HI);
nrbytes = (size_t) fs_m_in.REQ_NBYTES;
block_size = get_block_size( (dev_t) fs_m_in.REQ_DEV2);
rip.i_zone[0] = (zone_t) fs_m_in.REQ_DEV2;
rip.i_mode = I_BLOCK_SPECIAL;
rip.i_size = 0;
rdwt_err = OK; /* set to EIO if disk error occurs */
cum_io = 0;
/* Split the transfer into chunks that don't span two blocks. */
while (nrbytes > 0) {
off = rem64u(position, block_size); /* offset in blk*/
chunk = min(nrbytes, block_size - off);
/* Read or write 'chunk' bytes. */
r = rw_chunk(&rip, position, off, chunk, nrbytes, rw_flag, gid,
cum_io, block_size, &completed);
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
nrbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position = add64ul(position, chunk); /* position within the file */
}
fs_m_out.RES_SEEK_POS_LO = ex64lo(position);
fs_m_out.RES_SEEK_POS_HI = ex64hi(position);
if (rdwt_err != OK) r = rdwt_err; /* check for disk error */
if (rdwt_err == END_OF_FILE) r = OK;
fs_m_out.RES_NBYTES = cum_io;
return(r);
}
static __inline u32_t makehash(u32_t p1, u64_t p2)
{
u32_t offlo = ex64lo(p2), offhi = ex64hi(p2),
v = 0x12345678;
hash_mix(p1, offlo, offhi);
hash_final(offlo, offhi, v);
return v % HASHSIZE;
}
/*===========================================================================*
* do_read *
*===========================================================================*/
int do_read()
{
/* Read data from a file.
*/
struct inode *ino;
u64_t pos;
size_t count, size;
vir_bytes off;
char *ptr;
int r, chunk;
if ((ino = find_inode(m_in.REQ_INODE_NR)) == NULL)
return EINVAL;
if (IS_DIR(ino)) return EISDIR;
if ((r = get_handle(ino)) != OK)
return r;
pos = make64(m_in.REQ_SEEK_POS_LO, m_in.REQ_SEEK_POS_HI);
count = m_in.REQ_NBYTES;
assert(count > 0);
/* Use the buffer from below to eliminate extra copying. */
size = sffs_table->t_readbuf(&ptr);
off = 0;
while (count > 0) {
chunk = MIN(count, size);
if ((r = sffs_table->t_read(ino->i_file, ptr, chunk, pos)) <= 0)
break;
chunk = r;
r = sys_safecopyto(m_in.m_source, m_in.REQ_GRANT, off,
(vir_bytes) ptr, chunk, D);
if (r != OK)
break;
count -= chunk;
off += chunk;
pos = add64u(pos, chunk);
}
if (r < 0)
return r;
m_out.RES_SEEK_POS_HI = ex64hi(pos);
m_out.RES_SEEK_POS_LO = ex64lo(pos);
m_out.RES_NBYTES = off;
return OK;
}
static void testdiv(void)
{
u64_t q, r;
#if TIMED
struct timeval tvstart, tvend;
printf("i=0x%.8x%.8x; j=0x%.8x%.8x\n",
ex64hi(i), ex64lo(i),
ex64hi(j), ex64lo(j));
fflush(stdout);
if (gettimeofday(&tvstart, NULL) < 0) ERR;
#endif
/* division by zero has a separate test */
if (cmp64u(j, 0) == 0) {
testdiv0();
return;
}
/* perform division, store q in k to make ERR more informative */
q = div64(i, j);
r = rem64(i, j);
k = q;
#if TIMED
if (gettimeofday(&tvend, NULL) < 0) ERR;
tvend.tv_sec -= tvstart.tv_sec;
tvend.tv_usec -= tvstart.tv_usec;
if (tvend.tv_usec < 0) {
tvend.tv_sec -= 1;
tvend.tv_usec += 1000000;
}
printf("q=0x%.8x%.8x; r=0x%.8x%.8x; time=%d.%.6d\n",
ex64hi(q), ex64lo(q),
ex64hi(r), ex64lo(r),
tvend.tv_sec, tvend.tv_usec);
fflush(stdout);
#endif
/* compare to 64/32-bit division if possible */
if (!ex64hi(j)) {
if (cmp64(q, div64u64(i, ex64lo(j))) != 0) ERR;
if (!ex64hi(q)) {
if (cmp64u(q, div64u(i, ex64lo(j))) != 0) ERR;
}
if (cmp64u(r, rem64u(i, ex64lo(j))) != 0) ERR;
/* compare to 32-bit division if possible */
if (!ex64hi(i)) {
if (cmp64u(q, ex64lo(i) / ex64lo(j)) != 0) ERR;
if (cmp64u(r, ex64lo(i) % ex64lo(j)) != 0) ERR;
}
}
/* check results using i = q j + r and r < j */
if (cmp64(i, add64(mul64(q, j), r)) != 0) ERR;
if (cmp64(r, j) >= 0) ERR;
}
/*===========================================================================*
* do_lseek *
*===========================================================================*/
int do_lseek()
{
/* Perform the lseek(ls_fd, offset, whence) system call. */
register struct filp *rfilp;
int r = OK, seekfd, seekwhence;
off_t offset;
u64_t pos, newpos;
seekfd = job_m_in.ls_fd;
seekwhence = job_m_in.whence;
offset = (off_t) job_m_in.offset_lo;
/* Check to see if the file descriptor is valid. */
if ( (rfilp = get_filp(seekfd, VNODE_READ)) == NULL) return(err_code);
/* No lseek on pipes. */
if (S_ISFIFO(rfilp->filp_vno->v_mode)) {
unlock_filp(rfilp);
return(ESPIPE);
}
/* The value of 'whence' determines the start position to use. */
switch(seekwhence) {
case SEEK_SET: pos = cvu64(0); break;
case SEEK_CUR: pos = rfilp->filp_pos; break;
case SEEK_END: pos = cvul64(rfilp->filp_vno->v_size); break;
default: unlock_filp(rfilp); return(EINVAL);
}
if (offset >= 0)
newpos = add64ul(pos, offset);
else
newpos = sub64ul(pos, -offset);
/* Check for overflow. */
if (ex64hi(newpos) != 0) {
r = EOVERFLOW;
} else if ((off_t) ex64lo(newpos) < 0) { /* no negative file size */
r = EOVERFLOW;
} else {
/* insert the new position into the output message */
m_out.reply_l1 = ex64lo(newpos);
if (cmp64(newpos, rfilp->filp_pos) != 0) {
rfilp->filp_pos = newpos;
/* Inhibit read ahead request */
r = req_inhibread(rfilp->filp_vno->v_fs_e,
rfilp->filp_vno->v_inode_nr);
}
}
unlock_filp(rfilp);
return(r);
}
/*===========================================================================*
* do_readsuper *
*===========================================================================*/
PUBLIC int do_readsuper()
{
/* Mount the file system.
*/
char path[PATH_MAX];
struct inode *ino;
struct hgfs_attr attr;
int r;
dprintf(("HGFS: readsuper (dev %x, flags %x)\n",
(dev_t) m_in.REQ_DEV, m_in.REQ_FLAGS));
if (m_in.REQ_FLAGS & REQ_ISROOT) {
printf("HGFS: attempt to mount as root device\n");
return EINVAL;
}
state.read_only = !!(m_in.REQ_FLAGS & REQ_RDONLY);
state.dev = m_in.REQ_DEV;
init_dentry();
ino = init_inode();
attr.a_mask = HGFS_ATTR_MODE | HGFS_ATTR_SIZE;
/* We cannot continue if we fail to get the properties of the root inode at
* all, because we cannot guess the details of the root node to return to
* VFS. Print a (hopefully) helpful error message, and abort the mount.
*/
if ((r = verify_inode(ino, path, &attr)) != OK) {
if (r == EAGAIN)
printf("HGFS: shared folders disabled\n");
else if (opt.prefix[0] && (r == ENOENT || r == EACCES))
printf("HGFS: unable to access the given prefix directory\n");
else
printf("HGFS: unable to access shared folders\n");
return r;
}
m_out.RES_INODE_NR = INODE_NR(ino);
m_out.RES_MODE = get_mode(ino, attr.a_mode);
m_out.RES_FILE_SIZE_HI = ex64hi(attr.a_size);
m_out.RES_FILE_SIZE_LO = ex64lo(attr.a_size);
m_out.RES_UID = opt.uid;
m_out.RES_GID = opt.gid;
m_out.RES_DEV = NO_DEV;
m_out.RES_CONREQS = 1; /* We can handle only 1 request at a time */
state.mounted = TRUE;
return OK;
}
/*===========================================================================*
* scall_lseek *
*===========================================================================*/
int scall_lseek()
{
/* Perform the lseek(ls_fd, offset, whence) system call. */
register struct filp *rfilp;
int r;
long offset;
u64_t pos, newpos;
/* Check to see if the file descriptor is valid. */
if ( (rfilp = get_filp(m_in.ls_fd)) == NIL_FILP)
return(err_code);
/* No lseek on pipes. */
if (rfilp->filp_vno->v_pipe == I_PIPE)
return -ESPIPE;
/* The value of 'whence' determines the start position to use. */
switch(m_in.whence) {
case SEEK_SET:
pos = cvu64(0);
break;
case SEEK_CUR:
pos = rfilp->filp_pos;
break;
case SEEK_END:
pos = cvul64(rfilp->filp_vno->v_size);
break;
default:
return(-EINVAL);
}
offset= m_in.offset_lo;
if (offset >= 0)
newpos= add64ul(pos, offset);
else
newpos= sub64ul(pos, -offset);
/* Check for overflow. */
if (ex64hi(newpos) != 0)
return -EINVAL;
if (cmp64(newpos, rfilp->filp_pos) != 0) { /* Inhibit read ahead request */
r = req_inhibread(rfilp->filp_vno->v_fs_e, rfilp->filp_vno->v_inode_nr);
if (r != 0)
return r;
}
rfilp->filp_pos = newpos;
return ex64lo(newpos);
}
static void testmul(void)
{
int kdone, kidx;
u32_t ilo = ex64lo(i), jlo = ex64lo(j);
u64_t prod = mul64(i, j);
int prodbits;
/* compute maximum index of highest-order bit */
prodbits = bsr64(i) + bsr64(j) + 1;
if (cmp64u(i, 0) == 0 || cmp64u(j, 0) == 0) prodbits = -1;
if (bsr64(prod) > prodbits) ERR;
/* compare to 32-bit multiplication if possible */
if (ex64hi(i) == 0 && ex64hi(j) == 0) {
if (cmp64(prod, mul64u(ilo, jlo)) != 0) ERR;
/* if there is no overflow we can check against pure 32-bit */
if (prodbits < 32 && cmp64u(prod, ilo * jlo) != 0) ERR;
}
/* in 32-bit arith low-order DWORD matches regardless of overflow */
if (ex64lo(prod) != ilo * jlo) ERR;
/* multiplication by zero yields zero */
if (prodbits < 0 && cmp64u(prod, 0) != 0) ERR;
/* if there is no overflow, check absence of zero divisors */
if (prodbits >= 0 && prodbits < 64 && cmp64u(prod, 0) == 0) ERR;
/* commutativity */
if (cmp64(prod, mul64(j, i)) != 0) ERR;
/* loop though all argument value combinations for third argument */
for (kdone = 0, kidx = 0; k = getargval(kidx, &kdone), !kdone; kidx++) {
/* associativity */
if (cmp64(mul64(mul64(i, j), k), mul64(i, mul64(j, k))) != 0) ERR;
/* left and right distributivity */
if (cmp64(mul64(add64(i, j), k), add64(mul64(i, k), mul64(j, k))) != 0) ERR;
if (cmp64(mul64(i, add64(j, k)), add64(mul64(i, j), mul64(i, k))) != 0) ERR;
}
}
/*===========================================================================*
* print64 *
*===========================================================================*/
char *print64(u64_t p)
{
#define NB 10
static int n = 0;
static char buf[NB][100];
u32_t lo = ex64lo(p), hi = ex64hi(p);
n = (n+1) % NB;
if(!hi) sprintf(buf[n], "%lx", lo);
else sprintf(buf[n], "%lx%08lx", hi, lo);
return buf[n];
}
/*===========================================================================*
* fs_bread *
*===========================================================================*/
int fs_bread(void)
{
int r, rw_flag, chunk, block_size;
cp_grant_id_t gid;
int nrbytes;
u64_t position;
unsigned int off, cum_io;
int completed;
struct dir_record *dir;
r = OK;
rw_flag = (fs_m_in.m_type == REQ_BREAD ? READING : WRITING);
gid = fs_m_in.REQ_GRANT;
position = make64(fs_m_in.REQ_SEEK_POS_LO, fs_m_in.REQ_SEEK_POS_HI);
nrbytes = (unsigned) fs_m_in.REQ_NBYTES;
block_size = v_pri.logical_block_size_l;
dir = v_pri.dir_rec_root;
if(rw_flag == WRITING) return (EIO); /* Not supported */
rdwt_err = OK; /* set to EIO if disk error occurs */
cum_io = 0;
/* Split the transfer into chunks that don't span two blocks. */
while (nrbytes != 0) {
off = rem64u(position, block_size); /* offset in blk*/
chunk = MIN(nrbytes, block_size - off);
if (chunk < 0) chunk = block_size - off;
/* Read 'chunk' bytes. */
r = read_chunk(dir, position, off, chunk, (unsigned) nrbytes,
gid, cum_io, block_size, &completed);
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
nrbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position= add64ul(position, chunk); /* position within the file */
}
fs_m_out.RES_SEEK_POS_LO = ex64lo(position);
fs_m_out.RES_SEEK_POS_HI = ex64hi(position);
if (rdwt_err != OK) r = rdwt_err; /* check for disk error */
if (rdwt_err == END_OF_FILE) r = OK;
fs_m_out.RES_NBYTES = cum_io;
return(r);
}
/*===========================================================================*
* do_llseek *
*===========================================================================*/
PUBLIC int do_llseek()
{
/* Perform the llseek(ls_fd, offset, whence) system call. */
register struct filp *rfilp;
u64_t pos, newpos;
int r = OK;
/* Check to see if the file descriptor is valid. */
if ( (rfilp = get_filp(m_in.ls_fd, VNODE_READ)) == NULL) return(err_code);
/* No lseek on pipes. */
if (rfilp->filp_vno->v_pipe == I_PIPE) {
unlock_filp(rfilp);
return(ESPIPE);
}
/* The value of 'whence' determines the start position to use. */
switch(m_in.whence) {
case SEEK_SET: pos = cvu64(0); break;
case SEEK_CUR: pos = rfilp->filp_pos; break;
case SEEK_END: pos = cvul64(rfilp->filp_vno->v_size); break;
default: unlock_filp(rfilp); return(EINVAL);
}
newpos = add64(pos, make64(m_in.offset_lo, m_in.offset_high));
/* Check for overflow. */
if (( (long) m_in.offset_high > 0) && cmp64(newpos, pos) < 0)
r = EINVAL;
else if (( (long) m_in.offset_high < 0) && cmp64(newpos, pos) > 0)
r = EINVAL;
else {
rfilp->filp_pos = newpos;
/* insert the new position into the output message */
m_out.reply_l1 = ex64lo(newpos);
m_out.reply_l2 = ex64hi(newpos);
if (cmp64(newpos, rfilp->filp_pos) != 0) {
/* Inhibit read ahead request */
r = req_inhibread(rfilp->filp_vno->v_fs_e,
rfilp->filp_vno->v_inode_nr);
}
}
unlock_filp(rfilp);
return(r);
}
/*===========================================================================*
* fbd_transfer *
*===========================================================================*/
static int fbd_transfer(dev_t UNUSED(minor), int do_write, u64_t position,
endpoint_t endpt, iovec_t *iov, unsigned int nr_req, int flags)
{
/* Transfer data from or to the device. */
unsigned count;
size_t size, osize;
int i, hooks;
ssize_t r;
/* Compute the total size of the request. */
for (size = i = 0; i < nr_req; i++)
size += iov[i].iov_size;
osize = size;
count = nr_req;
hooks = rule_find(position, size,
do_write ? FBD_FLAG_WRITE : FBD_FLAG_READ);
#if DEBUG
printf("FBD: %s operation for pos %lx:%08lx size %u -> hooks %x\n",
do_write ? "write" : "read", ex64hi(position),
ex64lo(position), size, hooks);
#endif
if (hooks & PRE_HOOK)
rule_pre_hook(iov, &count, &size, &position);
if (count > 0) {
if (hooks & IO_HOOK) {
r = fbd_transfer_copy(do_write, position, endpt, iov,
count, size, flags);
} else {
r = fbd_transfer_direct(do_write, position, endpt, iov,
count, flags);
}
}
else r = 0;
if (hooks & POST_HOOK)
rule_post_hook(osize, &r);
#if DEBUG
printf("FBD: returning %d\n", r);
#endif
return r;
}
static double make_double(u64_t d)
{
/* Convert a 64-bit fixed point value into a double.
* This whole thing should be replaced by something better eventually.
*/
double value;
size_t i;
value = (double) ex64hi(d);
for (i = 0; i < sizeof(unsigned long); i += 2)
value *= 65536.0;
value += (double) ex64lo(d);
return value;
}
/*===========================================================================*
* vfs_request *
*===========================================================================*/
int vfs_request(int reqno, int fd, struct vmproc *vmp, u64_t offset, u32_t len,
vfs_callback_t reply_callback, void *cbarg, void *state, int statelen)
{
/* Perform an asynchronous request to VFS.
* We send a message of type VFS_VMCALL to VFS. VFS will respond
* with message type VM_VFS_REPLY. We send the request asynchronously
* and then handle the reply as it if were a VM_VFS_REPLY request.
*/
message *m;
static int reqid = 0;
struct vfs_request_node *reqnode;
reqid++;
assert(statelen <= STATELEN);
if(!SLABALLOC(reqnode)) {
printf("vfs_request: no memory for request node\n");
return ENOMEM;
}
m = &reqnode->reqmsg;
m->m_type = VFS_VMCALL;
m->VFS_VMCALL_REQ = reqno;
m->VFS_VMCALL_FD = fd;
m->VFS_VMCALL_REQID = reqid;
m->VFS_VMCALL_ENDPOINT = vmp->vm_endpoint;
m->VFS_VMCALL_OFFSET_LO = ex64lo(offset);
m->VFS_VMCALL_OFFSET_HI = ex64hi(offset);
m->VFS_VMCALL_LENGTH = len;
reqnode->who = vmp->vm_endpoint;
reqnode->req_id = reqid;
reqnode->next = first_queued;
reqnode->callback = reply_callback;
reqnode->opaque = cbarg;
if(state) memcpy(reqnode->reqstate, state, statelen);
first_queued = reqnode;
/* Send the request message if none pending. */
if(!active)
activate();
return OK;
}
/*===========================================================================*
* req_bpeek *
*===========================================================================*/
int req_bpeek(endpoint_t fs_e, dev_t dev, u64_t pos, unsigned int num_of_bytes)
{
message m;
memset(&m, 0, sizeof(m));
/* Fill in request message */
m.m_type = REQ_BPEEK;
m.REQ_DEV2 = dev;
m.REQ_SEEK_POS_LO = ex64lo(pos);
m.REQ_SEEK_POS_HI = ex64hi(pos);
m.REQ_NBYTES = num_of_bytes;
/* Send/rec request */
return fs_sendrec(fs_e, &m);
return(OK);
}
/*===========================================================================*
* get_range *
*===========================================================================*/
static size_t get_range(struct fbd_rule *rule, u64_t pos, size_t *size,
u64_t *skip)
{
/* Compute the range within the given request range that is affected
* by the given rule, and optionally the number of bytes preceding
* the range that are also affected by the rule.
*/
u64_t delta;
size_t off;
int to_eof;
to_eof = cmp64(rule->start, rule->end) >= 0;
if (cmp64(pos, rule->start) > 0) {
if (skip != NULL) *skip = sub64(pos, rule->start);
off = 0;
}
else {
if (skip != NULL) *skip = cvu64(0);
delta = sub64(rule->start, pos);
assert(ex64hi(delta) == 0);
off = ex64lo(delta);
}
if (!to_eof) {
assert(cmp64(pos, rule->end) < 0);
delta = sub64(rule->end, pos);
if (cmp64u(delta, *size) < 0)
*size = ex64lo(delta);
}
assert(*size > off);
*size -= off;
return off;
}
/*===========================================================================*
* fbd_transfer_direct *
*===========================================================================*/
static ssize_t fbd_transfer_direct(int do_write, u64_t position,
endpoint_t endpt, iovec_t *iov, unsigned int count, int flags)
{
/* Forward the entire transfer request, without any intervention. */
iovec_s_t iovec[NR_IOREQS];
cp_grant_id_t grant;
message m;
int i, r;
for (i = 0; i < count; i++) {
iovec[i].iov_size = iov[i].iov_size;
iovec[i].iov_grant = cpf_grant_indirect(driver_endpt, endpt,
iov[i].iov_addr);
assert(iovec[i].iov_grant != GRANT_INVALID);
}
grant = cpf_grant_direct(driver_endpt, (vir_bytes) iovec,
count * sizeof(iovec[0]), CPF_READ);
assert(grant != GRANT_INVALID);
m.m_type = do_write ? BDEV_SCATTER : BDEV_GATHER;
m.BDEV_MINOR = driver_minor;
m.BDEV_COUNT = count;
m.BDEV_GRANT = grant;
m.BDEV_FLAGS = flags;
m.BDEV_ID = 0;
m.BDEV_POS_LO = ex64lo(position);
m.BDEV_POS_HI = ex64hi(position);
if ((r = sendrec(driver_endpt, &m)) != OK)
panic("sendrec to driver failed (%d)\n", r);
if (m.m_type != BDEV_REPLY)
panic("invalid reply from driver (%d)\n", m.m_type);
cpf_revoke(grant);
for (i = 0; i < count; i++)
cpf_revoke(iovec[i].iov_grant);
return m.BDEV_STATUS;
}
static void dump_entry(btrace_entry *entry)
{
switch (entry->request) {
case BTREQ_OPEN: printf("OPEN"); break;
case BTREQ_CLOSE: printf("CLOSE"); break;
case BTREQ_READ: printf("READ"); break;
case BTREQ_WRITE: printf("WRITE"); break;
case BTREQ_GATHER: printf("GATHER"); break;
case BTREQ_SCATTER: printf("SCATTER"); break;
case BTREQ_IOCTL: printf("IOCTL"); break;
}
printf(" request\n");
switch (entry->request) {
case BTREQ_OPEN:
printf("- access:\t%x\n", entry->size);
break;
case BTREQ_READ:
case BTREQ_WRITE:
case BTREQ_GATHER:
case BTREQ_SCATTER:
printf("- position:\t%08lx%08lx\n",
ex64hi(entry->position), ex64lo(entry->position));
printf("- size:\t\t%u\n", entry->size);
printf("- flags:\t%x\n", entry->flags);
break;
case BTREQ_IOCTL:
printf("- request:\t%08x\n", entry->size);
break;
}
printf("- start:\t%u us\n", entry->start_time);
printf("- finish:\t%u us\n", entry->finish_time);
if (entry->result == BTRES_INPROGRESS)
printf("- result:\t(in progress)\n");
else
printf("- result:\t%d\n", entry->result);
printf("\n");
}
/*===========================================================================*
* req_breadwrite *
*===========================================================================*/
PUBLIC int req_breadwrite(
endpoint_t fs_e,
endpoint_t user_e,
dev_t dev,
u64_t pos,
unsigned int num_of_bytes,
char *user_addr,
int rw_flag,
u64_t *new_posp,
unsigned int *cum_iop
)
{
int r;
cp_grant_id_t grant_id;
message m;
grant_id = cpf_grant_magic(fs_e, user_e, (vir_bytes) user_addr, num_of_bytes,
(rw_flag == READING ? CPF_WRITE : CPF_READ));
if(grant_id == -1)
panic("req_breadwrite: cpf_grant_magic failed");
/* Fill in request message */
m.m_type = rw_flag == READING ? REQ_BREAD : REQ_BWRITE;
m.REQ_DEV2 = dev;
m.REQ_GRANT = grant_id;
m.REQ_SEEK_POS_LO = ex64lo(pos);
m.REQ_SEEK_POS_HI = ex64hi(pos);
m.REQ_NBYTES = num_of_bytes;
/* Send/rec request */
r = fs_sendrec(fs_e, &m);
cpf_revoke(grant_id);
if (r != OK) return(r);
/* Fill in response structure */
*new_posp = make64(m.RES_SEEK_POS_LO, m.RES_SEEK_POS_HI);
*cum_iop = m.RES_NBYTES;
return(OK);
}
PUBLIC void print_proc(struct proc *pp)
{
struct proc *depproc = NULL;
endpoint_t dep;
printf("%d: %s %d prio %d time %d/%d cycles 0x%x%08x cpu %2d "
"cr3 0x%lx rts %s misc %s sched %s ",
proc_nr(pp), pp->p_name, pp->p_endpoint,
pp->p_priority, pp->p_user_time,
pp->p_sys_time, ex64hi(pp->p_cycles),
ex64lo(pp->p_cycles), pp->p_cpu,
pp->p_seg.p_cr3,
rtsflagstr(pp->p_rts_flags), miscflagstr(pp->p_misc_flags),
schedulerstr(pp->p_scheduler));
print_sigmgr(pp);
dep = P_BLOCKEDON(pp);
if(dep != NONE) {
printf(" blocked on: ");
print_endpoint(dep);
}
printf("\n");
}
int startsim(void){
message m;
int i,j=0,piReady = -1;
u64_t cpuTimeDiff;
FILE *fp;
fp = fopen("/home/out","w");
for(i=0;i<HISTORY;i++){
pInfoPtrs[i] = (struct pi *) &pInfo[i][0];
}
for(i=0;i<HISTORY;i++){
pQhPtrs[i] = (struct qh*) &pQh[i][0];
}
/* Copy the pointer arrays so that you can go backward and forward through the history */
struct pi *pInfoPtrsCopy[HISTORY];
struct qh *pQhPtrsCopy[HISTORY];
for(i=0;i<HISTORY;i++){
pInfoPtrsCopy[i] = pInfoPtrs[i];
pQhPtrsCopy[i] = pQhPtrs[i];
}
m.m1_p1 = (char *) &pInfoPtrs;
m.m1_p2 = (char *) &piReady;
m.m1_i2 = SELF;
m.m1_i3 = HISTORY;
m.m2_p1 = (char *) &pQhPtrs;
m.m3_p1 = (char *) &cpuFreq;
int error = _syscall(PM_PROC_NR,STARTRECORD,&m);
procs();
for(j;j<HISTORY;j++){
while(piReady < j){
}
if(j==0){
fprintf(fp,"CPU frequency is: %lu Mhz\n",div64u(cpuFreq, 1000000));
}
printf("Simulation is %d%% complete.\n",(j*2)+2);
fprintf(fp,"Proc Table %d\n\n",j);
fprintf(fp,"Queue heads: ");
for(i=0;i<NR_SCHED_QUEUES;i++){
if(pQhPtrsCopy[j]->p_endpoint!=-1){
fprintf(fp,"Queue: %d %d ",i,pQhPtrsCopy[j]->p_endpoint);
}
pQhPtrsCopy[j]++;
}
fprintf(fp,"\n\n");
pQhPtrsCopy[j] = pQhPtrs[j];
/*Write out the runable queues in order */
for(i=0; i<NR_SCHED_QUEUES; i++){
fprintf(fp,"Priority Queue %d: ",i);
if(pQhPtrsCopy[j]->p_endpoint != -1){
printQ(pInfoPtrsCopy[j],pQhPtrsCopy[j]->p_endpoint,fp);
}
else{
fprintf(fp,"\n");
}
pQhPtrsCopy[j]++;
}
pQhPtrsCopy[j] = pQhPtrs[j]; /* Reset the Qh Pointers */
for (i=0;i<ALL_PROCS;i++){
if (!(pInfoPtrsCopy[j]->p_rts_flags == RTS_SLOT_FREE)){
if(j>0){
cpuTimeDiff = sub64(pInfoPtrsCopy[j]->p_cycles,pInfoPtrsCopy[j-1]->p_cycles);
}
fprintf(fp,"Process: %s, Endpoint: %d, Enter queue: %lu%lu, Time in Queue: %lu%lu, Dequeues: %lu, IPC Sync: %lu, IPC Async: %lu,Preempted: %lu,RTS: %x\n\t\t, Priority: %d, Next: %s Endpoint: %d, User time: %d, Sys Time: %d, CPU Cycles Elaps: %lu%lu\n",
pInfoPtrsCopy[j]->p_name,pInfoPtrsCopy[j]->p_endpoint,ex64hi(pInfoPtrsCopy[j]->p_times.enter_queue),ex64lo(pInfoPtrsCopy[j]->p_times.enter_queue),
ex64hi(pInfoPtrsCopy[j]->p_times.time_in_queue),ex64lo(pInfoPtrsCopy[j]->p_times.time_in_queue),
pInfoPtrsCopy[j]->p_times.dequeues,pInfoPtrsCopy[j]->p_times.ipc_sync,pInfoPtrsCopy[j]->p_times.ipc_async,
pInfoPtrsCopy[j]->p_times.preempted,pInfoPtrsCopy[j]->p_rts_flags,pInfoPtrsCopy[j]->p_priority, pInfoPtrsCopy[j]->p_nextready,
pInfoPtrsCopy[j]->p_nextready_endpoint,pInfoPtrsCopy[j]->p_user_time,pInfoPtrsCopy[j]->p_sys_time,ex64hi(cpuTimeDiff),
ex64lo(cpuTimeDiff));
}
pInfoPtrsCopy[j]++;
if(j>0){
pInfoPtrsCopy[j-1]++;
}
}
pInfoPtrsCopy[j] = pInfoPtrs[j];
}
m.m1_i3 = -1;
_syscall(PM_PROC_NR,STARTRECORD,&m);
return(0);
}
PUBLIC void context_stop(struct proc * p)
{
u64_t tsc, tsc_delta;
u64_t * __tsc_ctr_switch = get_cpulocal_var_ptr(tsc_ctr_switch);
#ifdef CONFIG_SMP
unsigned cpu = cpuid;
/*
* This function is called only if we switch from kernel to user or idle
* or back. Therefore this is a perfect location to place the big kernel
* lock which will hopefully disappear soon.
*
* If we stop accounting for KERNEL we must unlock the BKL. If account
* for IDLE we must not hold the lock
*/
if (p == proc_addr(KERNEL)) {
u64_t tmp;
read_tsc_64(&tsc);
tmp = sub64(tsc, *__tsc_ctr_switch);
kernel_ticks[cpu] = add64(kernel_ticks[cpu], tmp);
p->p_cycles = add64(p->p_cycles, tmp);
BKL_UNLOCK();
} else {
u64_t bkl_tsc;
atomic_t succ;
read_tsc_64(&bkl_tsc);
/* this only gives a good estimate */
succ = big_kernel_lock.val;
BKL_LOCK();
read_tsc_64(&tsc);
bkl_ticks[cpu] = add64(bkl_ticks[cpu], sub64(tsc, bkl_tsc));
bkl_tries[cpu]++;
bkl_succ[cpu] += !(!(succ == 0));
p->p_cycles = add64(p->p_cycles, sub64(tsc, *__tsc_ctr_switch));
#ifdef CONFIG_SMP
/*
* Since at the time we got a scheduling IPI we might have been
* waiting for BKL already, we may miss it due to a similar IPI to
* the cpu which is already waiting for us to handle its. This
* results in a live-lock of these two cpus.
*
* Therefore we always check if there is one pending and if so,
* we handle it straight away so the other cpu can continue and
* we do not deadlock.
*/
smp_sched_handler();
#endif
}
#else
read_tsc_64(&tsc);
p->p_cycles = add64(p->p_cycles, sub64(tsc, *__tsc_ctr_switch));
#endif
tsc_delta = sub64(tsc, *__tsc_ctr_switch);
if(kbill_ipc) {
kbill_ipc->p_kipc_cycles =
add64(kbill_ipc->p_kipc_cycles, tsc_delta);
kbill_ipc = NULL;
}
if(kbill_kcall) {
kbill_kcall->p_kcall_cycles =
add64(kbill_kcall->p_kcall_cycles, tsc_delta);
kbill_kcall = NULL;
}
/*
* deduct the just consumed cpu cycles from the cpu time left for this
* process during its current quantum. Skip IDLE and other pseudo kernel
* tasks
*/
if (p->p_endpoint >= 0) {
#if DEBUG_RACE
make_zero64(p->p_cpu_time_left);
#else
/* if (tsc_delta < p->p_cpu_time_left) in 64bit */
if (ex64hi(tsc_delta) < ex64hi(p->p_cpu_time_left) ||
(ex64hi(tsc_delta) == ex64hi(p->p_cpu_time_left) &&
ex64lo(tsc_delta) < ex64lo(p->p_cpu_time_left)))
p->p_cpu_time_left = sub64(p->p_cpu_time_left, tsc_delta);
else {
make_zero64(p->p_cpu_time_left);
}
#endif
}
*__tsc_ctr_switch = tsc;
}
/*===========================================================================*
* do_fcntl *
*===========================================================================*/
PUBLIC int do_fcntl()
{
/* Perform the fcntl(fd, request, ...) system call. */
register struct filp *f;
int new_fd, r, fl;
struct filp *dummy;
/* Is the file descriptor valid? */
if ((f = get_filp(m_in.fd)) == NULL) return(err_code);
switch (m_in.request) {
case F_DUPFD:
/* This replaces the old dup() system call. */
if (m_in.addr < 0 || m_in.addr >= OPEN_MAX) return(EINVAL);
if ((r = get_fd(m_in.addr, 0, &new_fd, &dummy)) != OK) return(r);
f->filp_count++;
fp->fp_filp[new_fd] = f;
return(new_fd);
case F_GETFD:
/* Get close-on-exec flag (FD_CLOEXEC in POSIX Table 6-2). */
return( FD_ISSET(m_in.fd, &fp->fp_cloexec_set) ? FD_CLOEXEC : 0);
case F_SETFD:
/* Set close-on-exec flag (FD_CLOEXEC in POSIX Table 6-2). */
if(m_in.addr & FD_CLOEXEC)
FD_SET(m_in.fd, &fp->fp_cloexec_set);
else
FD_CLR(m_in.fd, &fp->fp_cloexec_set);
return(OK);
case F_GETFL:
/* Get file status flags (O_NONBLOCK and O_APPEND). */
fl = f->filp_flags & (O_NONBLOCK | O_APPEND | O_ACCMODE);
return(fl);
case F_SETFL:
/* Set file status flags (O_NONBLOCK and O_APPEND). */
fl = O_NONBLOCK | O_APPEND | O_REOPEN;
f->filp_flags = (f->filp_flags & ~fl) | (m_in.addr & fl);
return(OK);
case F_GETLK:
case F_SETLK:
case F_SETLKW:
/* Set or clear a file lock. */
r = lock_op(f, m_in.request);
return(r);
case F_FREESP:
{
/* Free a section of a file. Preparation is done here, actual freeing
* in freesp_inode().
*/
off_t start, end;
struct flock flock_arg;
signed long offset;
/* Check if it's a regular file. */
if((f->filp_vno->v_mode & I_TYPE) != I_REGULAR) return(EINVAL);
if (!(f->filp_mode & W_BIT)) return(EBADF);
/* Copy flock data from userspace. */
if((r = sys_datacopy(who_e, (vir_bytes) m_in.name1, SELF,
(vir_bytes) &flock_arg, (phys_bytes) sizeof(flock_arg))) != OK)
return(r);
/* Convert starting offset to signed. */
offset = (signed long) flock_arg.l_start;
/* Figure out starting position base. */
switch(flock_arg.l_whence) {
case SEEK_SET: start = 0; break;
case SEEK_CUR:
if (ex64hi(f->filp_pos) != 0)
panic("do_fcntl: position in file too high");
start = ex64lo(f->filp_pos);
break;
case SEEK_END: start = f->filp_vno->v_size; break;
default: return EINVAL;
}
/* Check for overflow or underflow. */
if(offset > 0 && start + offset < start) return EINVAL;
if(offset < 0 && start + offset > start) return EINVAL;
start += offset;
if(start < 0) return EINVAL;
if(flock_arg.l_len != 0) {
if(start >= f->filp_vno->v_size) return EINVAL;
end = start + flock_arg.l_len;
if(end <= start) return EINVAL;
if(end > f->filp_vno->v_size) end = f->filp_vno->v_size;
} else {
end = 0;
}
r = req_ftrunc(f->filp_vno->v_fs_e, f->filp_vno->v_inode_nr, start,
end);
if(r == OK && flock_arg.l_len == 0)
f->filp_vno->v_size = start;
return(r);
}
default:
return(EINVAL);
}
}
/*===========================================================================*
* m_transfer *
*===========================================================================*/
static int m_transfer(
endpoint_t endpt, /* endpoint of grant owner */
int opcode, /* DEV_GATHER_S or DEV_SCATTER_S */
u64_t pos64, /* offset on device to read or write */
iovec_t *iov, /* pointer to read or write request vector */
unsigned int nr_req, /* length of request vector */
endpoint_t UNUSED(user_endpt),/* endpoint of user process */
unsigned int UNUSED(flags)
)
{
/* Read or write one the driver's character devices. */
unsigned count, left, chunk;
vir_bytes vir_offset = 0;
struct device *dv;
unsigned long dv_size;
int s, r;
off_t position;
cp_grant_id_t grant;
vir_bytes dev_vaddr;
/* ZERO_DEV and NULL_DEV are infinite in size. */
if (m_device != ZERO_DEV && m_device != NULL_DEV && ex64hi(pos64) != 0)
return OK; /* Beyond EOF */
position= cv64ul(pos64);
/* Get minor device number and check for /dev/null. */
dv = &m_geom[m_device];
dv_size = cv64ul(dv->dv_size);
dev_vaddr = m_vaddrs[m_device];
while (nr_req > 0) {
/* How much to transfer and where to / from. */
count = iov->iov_size;
grant = (cp_grant_id_t) iov->iov_addr;
switch (m_device) {
/* No copying; ignore request. */
case NULL_DEV:
if (opcode == DEV_GATHER_S) return(OK); /* always at EOF */
break;
/* Virtual copying. For kernel memory. */
default:
case KMEM_DEV:
if(!dev_vaddr || dev_vaddr == (vir_bytes) MAP_FAILED) {
printf("MEM: dev %d not initialized\n", m_device);
return EIO;
}
if (position >= dv_size) return(OK); /* check for EOF */
if (position + count > dv_size) count = dv_size - position;
if (opcode == DEV_GATHER_S) { /* copy actual data */
r=sys_safecopyto(endpt, grant, vir_offset,
dev_vaddr + position, count);
} else {
r=sys_safecopyfrom(endpt, grant, vir_offset,
dev_vaddr + position, count);
}
if(r != OK) {
panic("I/O copy failed: %d", r);
}
break;
/* Physical copying. Only used to access entire memory.
* Transfer one 'page window' at a time.
*/
case MEM_DEV:
{
u32_t pagestart, page_off;
static u32_t pagestart_mapped;
static int any_mapped = 0;
static char *vaddr;
int r;
u32_t subcount;
phys_bytes mem_phys;
if (position >= dv_size)
return(OK); /* check for EOF */
if (position + count > dv_size)
count = dv_size - position;
mem_phys = position;
page_off = mem_phys % PAGE_SIZE;
pagestart = mem_phys - page_off;
/* All memory to the map call has to be page-aligned.
* Don't have to map same page over and over.
*/
if(!any_mapped || pagestart_mapped != pagestart) {
if(any_mapped) {
if(vm_unmap_phys(SELF, vaddr, PAGE_SIZE) != OK)
panic("vm_unmap_phys failed");
any_mapped = 0;
}
vaddr = vm_map_phys(SELF, (void *) pagestart, PAGE_SIZE);
if(vaddr == MAP_FAILED)
r = ENOMEM;
else
r = OK;
if(r != OK) {
printf("memory: vm_map_phys failed\n");
return r;
}
any_mapped = 1;
pagestart_mapped = pagestart;
}
/* how much to be done within this page. */
subcount = PAGE_SIZE-page_off;
if(subcount > count)
subcount = count;
if (opcode == DEV_GATHER_S) { /* copy data */
s=sys_safecopyto(endpt, grant,
vir_offset, (vir_bytes) vaddr+page_off, subcount);
} else {
s=sys_safecopyfrom(endpt, grant,
vir_offset, (vir_bytes) vaddr+page_off, subcount);
}
if(s != OK)
return s;
count = subcount;
break;
}
/* Null byte stream generator. */
case ZERO_DEV:
if (opcode == DEV_GATHER_S) {
size_t suboffset = 0;
left = count;
while (left > 0) {
chunk = (left > ZERO_BUF_SIZE) ? ZERO_BUF_SIZE : left;
s=sys_safecopyto(endpt, grant,
vir_offset+suboffset, (vir_bytes) dev_zero, chunk);
if(s != OK)
return s;
left -= chunk;
suboffset += chunk;
}
}
break;
}
/* Book the number of bytes transferred. */
position += count;
vir_offset += count;
if ((iov->iov_size -= count) == 0) { iov++; nr_req--; vir_offset = 0; }
}
return(OK);
}
/*===========================================================================*
* do_fcntl *
*===========================================================================*/
int do_fcntl()
{
/* Perform the fcntl(fd, request, ...) system call. */
register struct filp *f;
int new_fd, fl, r = OK, fcntl_req, fcntl_argx;
tll_access_t locktype;
scratch(fp).file.fd_nr = job_m_in.fd;
scratch(fp).io.io_buffer = job_m_in.buffer;
scratch(fp).io.io_nbytes = job_m_in.nbytes; /* a.k.a. m_in.request */
fcntl_req = job_m_in.request;
fcntl_argx = job_m_in.addr;
/* Is the file descriptor valid? */
locktype = (fcntl_req == F_FREESP) ? VNODE_WRITE : VNODE_READ;
if ((f = get_filp(scratch(fp).file.fd_nr, locktype)) == NULL)
return(err_code);
switch (fcntl_req) {
case F_DUPFD:
/* This replaces the old dup() system call. */
if (fcntl_argx < 0 || fcntl_argx >= OPEN_MAX) r = EINVAL;
else if ((r = get_fd(fcntl_argx, 0, &new_fd, NULL)) == OK) {
f->filp_count++;
fp->fp_filp[new_fd] = f;
FD_SET(new_fd, &fp->fp_filp_inuse);
r = new_fd;
}
break;
case F_GETFD:
/* Get close-on-exec flag (FD_CLOEXEC in POSIX Table 6-2). */
r = 0;
if (FD_ISSET(scratch(fp).file.fd_nr, &fp->fp_cloexec_set))
r = FD_CLOEXEC;
break;
case F_SETFD:
/* Set close-on-exec flag (FD_CLOEXEC in POSIX Table 6-2). */
if (fcntl_argx & FD_CLOEXEC)
FD_SET(scratch(fp).file.fd_nr, &fp->fp_cloexec_set);
else
FD_CLR(scratch(fp).file.fd_nr, &fp->fp_cloexec_set);
break;
case F_GETFL:
/* Get file status flags (O_NONBLOCK and O_APPEND). */
fl = f->filp_flags & (O_NONBLOCK | O_APPEND | O_ACCMODE);
r = fl;
break;
case F_SETFL:
/* Set file status flags (O_NONBLOCK and O_APPEND). */
fl = O_NONBLOCK | O_APPEND | O_REOPEN;
f->filp_flags = (f->filp_flags & ~fl) | (fcntl_argx & fl);
break;
case F_GETLK:
case F_SETLK:
case F_SETLKW:
/* Set or clear a file lock. */
r = lock_op(f, fcntl_req);
break;
case F_FREESP:
{
/* Free a section of a file */
off_t start, end;
struct flock flock_arg;
signed long offset;
/* Check if it's a regular file. */
if (!S_ISREG(f->filp_vno->v_mode)) r = EINVAL;
else if (!(f->filp_mode & W_BIT)) r = EBADF;
else
/* Copy flock data from userspace. */
r = sys_datacopy(who_e, (vir_bytes) scratch(fp).io.io_buffer,
SELF, (vir_bytes) &flock_arg,
sizeof(flock_arg));
if (r != OK) break;
/* Convert starting offset to signed. */
offset = (signed long) flock_arg.l_start;
/* Figure out starting position base. */
switch(flock_arg.l_whence) {
case SEEK_SET: start = 0; break;
case SEEK_CUR:
if (ex64hi(f->filp_pos) != 0)
panic("do_fcntl: position in file too high");
start = ex64lo(f->filp_pos);
break;
case SEEK_END: start = f->filp_vno->v_size; break;
default: r = EINVAL;
}
if (r != OK) break;
/* Check for overflow or underflow. */
if (offset > 0 && start + offset < start) r = EINVAL;
else if (offset < 0 && start + offset > start) r = EINVAL;
else {
start += offset;
if (start < 0) r = EINVAL;
}
if (r != OK) break;
if (flock_arg.l_len != 0) {
if (start >= f->filp_vno->v_size) r = EINVAL;
else if ((end = start + flock_arg.l_len) <= start) r = EINVAL;
else if (end > f->filp_vno->v_size) end = f->filp_vno->v_size;
} else {
end = 0;
}
if (r != OK) break;
r = req_ftrunc(f->filp_vno->v_fs_e, f->filp_vno->v_inode_nr,start,end);
if (r == OK && flock_arg.l_len == 0)
f->filp_vno->v_size = start;
break;
}
case F_GETNOSIGPIPE:
/* POSIX: return value other than -1 is flag is set, else -1 */
r = -1;
if (f->filp_flags & O_NOSIGPIPE)
r = 0;
break;
case F_SETNOSIGPIPE:
fl = (O_NOSIGPIPE);
f->filp_flags = (f->filp_flags & ~fl) | (fcntl_argx & fl);
break;
default:
r = EINVAL;
}
unlock_filp(f);
return(r);
}
/*===========================================================================*
* m_block_transfer *
*===========================================================================*/
static int m_block_transfer(
dev_t minor, /* minor device number */
int do_write, /* read or write? */
u64_t pos64, /* offset on device to read or write */
endpoint_t endpt, /* process doing the request */
iovec_t *iov, /* pointer to read or write request vector */
unsigned int nr_req, /* length of request vector */
int UNUSED(flags) /* transfer flags */
)
{
/* Read or write one the driver's block devices. */
unsigned count;
vir_bytes vir_offset = 0;
struct device *dv;
unsigned long dv_size;
int r;
off_t position;
vir_bytes dev_vaddr;
cp_grant_id_t grant;
ssize_t total = 0;
/* Get minor device information. */
if ((dv = m_block_part(minor)) == NULL) return(ENXIO);
dv_size = cv64ul(dv->dv_size);
dev_vaddr = m_vaddrs[minor];
if (ex64hi(pos64) != 0)
return OK; /* Beyond EOF */
position= cv64ul(pos64);
while (nr_req > 0) {
/* How much to transfer and where to / from. */
count = iov->iov_size;
grant = (cp_grant_id_t) iov->iov_addr;
/* Virtual copying. For RAM disks and internal FS. */
if(!dev_vaddr || dev_vaddr == (vir_bytes) MAP_FAILED) {
printf("MEM: dev %d not initialized\n", minor);
return EIO;
}
if (position >= dv_size) return(total); /* check for EOF */
if (position + count > dv_size) count = dv_size - position;
if (!do_write) { /* copy actual data */
r=sys_safecopyto(endpt, grant, vir_offset,
dev_vaddr + position, count);
} else {
r=sys_safecopyfrom(endpt, grant, vir_offset,
dev_vaddr + position, count);
}
if(r != OK) {
panic("I/O copy failed: %d", r);
}
/* Book the number of bytes transferred. */
position += count;
vir_offset += count;
total += count;
if ((iov->iov_size -= count) == 0) { iov++; nr_req--; vir_offset = 0; }
}
return(total);
}
/*===========================================================================*
* m_block_ioctl *
*===========================================================================*/
static int m_block_ioctl(dev_t minor, unsigned int request, endpoint_t endpt,
cp_grant_id_t grant)
{
/* I/O controls for the block devices of the memory driver. Currently there is
* one I/O control specific to the memory driver:
* - MIOCRAMSIZE: to set the size of the RAM disk.
*/
struct device *dv;
u32_t ramdev_size;
int s;
void *mem;
if (request != MIOCRAMSIZE)
return EINVAL;
/* Someone wants to create a new RAM disk with the given size.
* A ramdisk can be created only once, and only on RAM disk device.
*/
if ((dv = m_block_part(minor)) == NULL) return ENXIO;
if((minor < RAM_DEV_FIRST || minor > RAM_DEV_LAST) && minor != RAM_DEV_OLD) {
printf("MEM: MIOCRAMSIZE: %d not a ramdisk\n", minor);
return EINVAL;
}
/* Get request structure */
s= sys_safecopyfrom(endpt, grant, 0, (vir_bytes)&ramdev_size,
sizeof(ramdev_size));
if (s != OK)
return s;
if(m_vaddrs[minor] && !cmp64(dv->dv_size, cvul64(ramdev_size))) {
return(OK);
}
/* openct is 1 for the ioctl(). */
if(openct[minor] != 1) {
printf("MEM: MIOCRAMSIZE: %d in use (count %d)\n",
minor, openct[minor]);
return(EBUSY);
}
if(m_vaddrs[minor]) {
u32_t size;
if(ex64hi(dv->dv_size)) {
panic("huge old ramdisk");
}
size = ex64lo(dv->dv_size);
minix_munmap((void *) m_vaddrs[minor], size);
m_vaddrs[minor] = (vir_bytes) NULL;
}
#if DEBUG
printf("MEM:%d: allocating ramdisk of size 0x%x\n", minor, ramdev_size);
#endif
/* Try to allocate a piece of memory for the RAM disk. */
if((mem = minix_mmap(NULL, ramdev_size, PROT_READ|PROT_WRITE,
MAP_PREALLOC|MAP_ANON, -1, 0)) == MAP_FAILED) {
printf("MEM: failed to get memory for ramdisk\n");
return(ENOMEM);
}
m_vaddrs[minor] = (vir_bytes) mem;
dv->dv_size = cvul64(ramdev_size);
return(OK);
}
/*===========================================================================*
* fs_breadwrite_o *
*===========================================================================*/
PUBLIC int fs_breadwrite_o(void)
{
int r, usr, rw_flag, chunk, block_size;
int nrbytes;
u64_t position;
unsigned int off, cum_io;
mode_t mode_word;
int completed, r2 = OK;
char *user_addr;
/* Pseudo inode for rw_chunk */
struct inode rip;
r = OK;
/* Get the values from the request message */
rw_flag = (fs_m_in.m_type == REQ_BREAD_O ? READING : WRITING);
usr = fs_m_in.REQ_XFD_WHO_E;
position = make64(fs_m_in.REQ_XFD_POS_LO, fs_m_in.REQ_XFD_POS_HI);
nrbytes = (unsigned) fs_m_in.REQ_XFD_NBYTES;
user_addr = fs_m_in.REQ_XFD_USER_ADDR;
block_size = get_block_size(fs_m_in.REQ_XFD_BDEV);
rip.i_zone[0] = fs_m_in.REQ_XFD_BDEV;
rip.i_mode = I_BLOCK_SPECIAL;
rip.i_size = 0;
rdwt_err = OK; /* set to EIO if disk error occurs */
cum_io = 0;
/* Split the transfer into chunks that don't span two blocks. */
while (nrbytes != 0) {
off = rem64u(position, block_size); /* offset in blk*/
chunk = MIN(nrbytes, block_size - off);
if (chunk < 0) chunk = block_size - off;
/* Read or write 'chunk' bytes. */
r = rw_chunk(&rip, position, off, chunk, (unsigned) nrbytes,
rw_flag, user_addr, D, usr, block_size, &completed);
if (r != OK) break; /* EOF reached */
if (rdwt_err < 0) break;
/* Update counters and pointers. */
user_addr += chunk; /* user buffer address */
nrbytes -= chunk; /* bytes yet to be read */
cum_io += chunk; /* bytes read so far */
position= add64ul(position, chunk); /* position within the file */
}
fs_m_out.RES_XFD_POS_LO = ex64lo(position);
fs_m_out.RES_XFD_POS_HI = ex64hi(position);
if (rdwt_err != OK) r = rdwt_err; /* check for disk error */
if (rdwt_err == END_OF_FILE) r = OK;
fs_m_out.RES_XFD_CUM_IO = cum_io;
return(r);
}
/*===========================================================================*
* read_write *
*===========================================================================*/
int read_write(u64_t pos, char *bufa, char *bufb, size_t *sizep, int request)
{
iovec_s_t vectors[2][NR_IOREQS];
message m1, m2;
cp_grant_id_t gids[2];
int r, both, count;
gids[0] = gids[1] = GRANT_INVALID;
/* Send two requests only if mirroring is enabled and the given request
* is either FLT_READ2 or FLT_WRITE.
*/
both = (USE_MIRROR && request != FLT_READ);
count = paired_grant(bufa, bufb, request, gids, vectors, *sizep, both);
memset(&m1, 0, sizeof(m1));
m1.m_type = (request == FLT_WRITE) ? BDEV_SCATTER : BDEV_GATHER;
m1.BDEV_COUNT = count;
m1.BDEV_POS_LO = ex64lo(pos);
m1.BDEV_POS_HI = ex64hi(pos);
m2 = m1;
m1.BDEV_GRANT = gids[0];
m2.BDEV_GRANT = gids[1];
r = paired_sendrec(&m1, &m2, both);
paired_revoke(gids, vectors, count, both);
if(r != OK) {
#if DEBUG
if (r != RET_REDO)
printf("Filter: paired_sendrec returned %d\n", r);
#endif
return r;
}
if (m1.m_type != BDEV_REPLY || m1.BDEV_STATUS < 0) {
printf("Filter: unexpected/invalid reply from main driver: "
"(%x, %d)\n", m1.m_type, m1.BDEV_STATUS);
return bad_driver(DRIVER_MAIN, BD_PROTO,
(m1.m_type == BDEV_REPLY) ? m1.BDEV_STATUS : EFAULT);
}
if (m1.BDEV_STATUS != (ssize_t) *sizep) {
printf("Filter: truncated reply from main driver\n");
/* If the driver returned a value *larger* than we requested,
* OR if we did NOT exceed the disk size, then we should
* report the driver for acting strangely!
*/
if (m1.BDEV_STATUS > (ssize_t) *sizep ||
cmp64(add64u(pos, m1.BDEV_STATUS), disk_size) < 0)
return bad_driver(DRIVER_MAIN, BD_PROTO, EFAULT);
/* Return the actual size. */
*sizep = m1.BDEV_STATUS;
}
if (both) {
if (m2.m_type != BDEV_REPLY || m2.BDEV_STATUS < 0) {
printf("Filter: unexpected/invalid reply from "
"backup driver (%x, %d)\n",
m2.m_type, m2.BDEV_STATUS);
return bad_driver(DRIVER_BACKUP, BD_PROTO,
m2.m_type == BDEV_REPLY ? m2.BDEV_STATUS :
EFAULT);
}
if (m2.BDEV_STATUS != (ssize_t) *sizep) {
printf("Filter: truncated reply from backup driver\n");
/* As above */
if (m2.BDEV_STATUS > (ssize_t) *sizep ||
cmp64(add64u(pos, m2.BDEV_STATUS),
disk_size) < 0)
return bad_driver(DRIVER_BACKUP, BD_PROTO,
EFAULT);
/* Return the actual size. */
if ((ssize_t) *sizep >= m2.BDEV_STATUS)
*sizep = m2.BDEV_STATUS;
}
}
return OK;
}
/*===========================================================================*
* do_create *
*===========================================================================*/
PUBLIC int do_create()
{
/* Create a new file.
*/
char path[PATH_MAX], name[NAME_MAX+1];
struct inode *parent, *ino;
struct hgfs_attr attr;
hgfs_file_t handle;
int r;
/* We cannot create files on a read-only file system. */
if (state.read_only)
return EROFS;
/* Get path, name, parent inode and possibly inode for the given path. */
if ((r = get_name(m_in.REQ_GRANT, m_in.REQ_PATH_LEN, name)) != OK)
return r;
if (!strcmp(name, ".") || !strcmp(name, "..")) return EEXIST;
if ((parent = find_inode(m_in.REQ_INODE_NR)) == NULL)
return EINVAL;
if ((r = verify_dentry(parent, name, path, &ino)) != OK)
return r;
/* Are we going to need a new inode upon success?
* Then make sure there is one available before trying anything.
*/
if (ino == NULL || ino->i_ref > 1 || HAS_CHILDREN(ino)) {
if (!have_free_inode()) {
if (ino != NULL)
put_inode(ino);
return ENFILE;
}
}
/* Perform the actual create call. */
r = hgfs_open(path, O_CREAT | O_EXCL | O_RDWR, m_in.REQ_MODE, &handle);
if (r != OK) {
/* Let's not try to be too clever with error codes here. If something
* is wrong with the directory, we'll find out later anyway.
*/
if (ino != NULL)
put_inode(ino);
return r;
}
/* Get the created file's attributes. */
attr.a_mask = HGFS_ATTR_MODE | HGFS_ATTR_SIZE;
r = hgfs_getattr(path, &attr);
/* If this fails, or returns a directory, we have a problem. This
* scenario is in fact possible with race conditions.
* Simulate a close and return a somewhat appropriate error.
*/
if (r != OK || S_ISDIR(attr.a_mode)) {
printf("HGFS: lost file after creation!\n");
hgfs_close(handle);
if (ino != NULL) {
del_dentry(ino);
put_inode(ino);
}
return (r == OK) ? EEXIST : r;
}
/* We do assume that the HGFS open(O_CREAT|O_EXCL) did its job.
* If we previousy found an inode, get rid of it now. It's old.
*/
if (ino != NULL) {
del_dentry(ino);
put_inode(ino);
}
/* Associate the open file handle with an inode, and reply with its details.
*/
ino = get_free_inode();
assert(ino != NULL); /* we checked before whether we had a free one */
ino->i_file = handle;
ino->i_flags = I_HANDLE;
add_dentry(parent, name, ino);
m_out.RES_INODE_NR = INODE_NR(ino);
m_out.RES_MODE = get_mode(ino, attr.a_mode);
m_out.RES_FILE_SIZE_HI = ex64hi(attr.a_size);
m_out.RES_FILE_SIZE_LO = ex64lo(attr.a_size);
m_out.RES_UID = opt.uid;
m_out.RES_GID = opt.gid;
m_out.RES_DEV = NO_DEV;
return OK;
}