int
procfs_doprocstatus(PFS_FILL_ARGS)
{
struct session *sess;
struct thread *tdfirst;
struct tty *tp;
struct ucred *cr;
const char *wmesg;
char *pc;
char *sep;
int pid, ppid, pgid, sid;
int i;
pid = p->p_pid;
PROC_LOCK(p);
ppid = p->p_pptr ? p->p_pptr->p_pid : 0;
pgid = p->p_pgrp->pg_id;
sess = p->p_pgrp->pg_session;
SESS_LOCK(sess);
sid = sess->s_leader ? sess->s_leader->p_pid : 0;
/* comm pid ppid pgid sid tty ctty,sldr start ut st wmsg
euid ruid rgid,egid,groups[1 .. ngroups]
*/
pc = p->p_comm;
do {
if (*pc < 33 || *pc > 126 || *pc == '\\')
sbuf_printf(sb, "\\%03o", *pc);
else
sbuf_putc(sb, *pc);
} while (*++pc);
sbuf_printf(sb, " %d %d %d %d ", pid, ppid, pgid, sid);
if ((p->p_flag & P_CONTROLT) && (tp = sess->s_ttyp))
sbuf_printf(sb, "%s ", devtoname(tp->t_dev));
else
sbuf_printf(sb, "- ");
sep = "";
if (sess->s_ttyvp) {
sbuf_printf(sb, "%sctty", sep);
sep = ",";
}
if (SESS_LEADER(p)) {
sbuf_printf(sb, "%ssldr", sep);
sep = ",";
}
SESS_UNLOCK(sess);
if (*sep != ',') {
sbuf_printf(sb, "noflags");
}
tdfirst = FIRST_THREAD_IN_PROC(p);
thread_lock(tdfirst);
if (tdfirst->td_wchan != NULL) {
KASSERT(tdfirst->td_wmesg != NULL,
("wchan %p has no wmesg", tdfirst->td_wchan));
wmesg = tdfirst->td_wmesg;
} else
wmesg = "nochan";
thread_unlock(tdfirst);
if (p->p_flag & P_INMEM) {
struct timeval start, ut, st;
PROC_SLOCK(p);
calcru(p, &ut, &st);
PROC_SUNLOCK(p);
start = p->p_stats->p_start;
timevaladd(&start, &boottime);
sbuf_printf(sb, " %jd,%ld %jd,%ld %jd,%ld",
(intmax_t)start.tv_sec, start.tv_usec,
(intmax_t)ut.tv_sec, ut.tv_usec,
(intmax_t)st.tv_sec, st.tv_usec);
} else
sbuf_printf(sb, " -1,-1 -1,-1 -1,-1");
sbuf_printf(sb, " %s", wmesg);
cr = p->p_ucred;
sbuf_printf(sb, " %lu %lu %lu",
(u_long)cr->cr_uid,
(u_long)cr->cr_ruid,
(u_long)cr->cr_rgid);
/* egid (cr->cr_svgid) is equal to cr_ngroups[0]
see also getegid(2) in /sys/kern/kern_prot.c */
for (i = 0; i < cr->cr_ngroups; i++) {
sbuf_printf(sb, ",%lu", (u_long)cr->cr_groups[i]);
}
if (jailed(cr)) {
mtx_lock(&cr->cr_prison->pr_mtx);
sbuf_printf(sb, " %s",
prison_name(td->td_ucred->cr_prison, cr->cr_prison));
mtx_unlock(&cr->cr_prison->pr_mtx);
} else {
sbuf_printf(sb, " -");
}
PROC_UNLOCK(p);
sbuf_printf(sb, "\n");
return (0);
}
/*
* Handle the Specialix ioctls on the control dev.
*/
static int
si_Sioctl(struct cdev *dev, u_long cmd, caddr_t data, int flag, struct thread *td)
{
struct si_softc *xsc;
struct si_port *xpp;
volatile struct si_reg *regp;
struct si_tcsi *dp;
struct si_pstat *sps;
int *ip, error = 0;
int oldspl;
int card, port;
DPRINT((0, DBG_ENTRY|DBG_IOCTL, "si_Sioctl(%s,0x%lx,0x%x)\n",
devtoname(dev), cmd, data));
mtx_assert(&Giant, MA_OWNED);
#if 1
DPRINT((0, DBG_IOCTL, "TCSI_PORT=%x\n", TCSI_PORT));
DPRINT((0, DBG_IOCTL, "TCSI_CCB=%x\n", TCSI_CCB));
#endif
oldspl = spltty(); /* better safe than sorry */
ip = (int *)data;
#define SUCHECK if ((error = priv_check(td, PRIV_DRIVER))) goto out
switch (cmd) {
case TCSIPORTS:
*ip = si_Nports;
goto out;
case TCSIMODULES:
*ip = si_Nmodules;
goto out;
case TCSISDBG_ALL:
SUCHECK;
si_debug = *ip;
goto out;
case TCSIGDBG_ALL:
*ip = si_debug;
goto out;
default:
/*
* Check that a controller for this port exists
*/
/* may also be a struct si_pstat, a superset of si_tcsi */
dp = (struct si_tcsi *)data;
sps = (struct si_pstat *)data;
card = dp->tc_card;
xsc = devclass_get_softc(si_devclass, card); /* check.. */
if (xsc == NULL || xsc->sc_type == SIEMPTY) {
error = ENOENT;
goto out;
}
/*
* And check that a port exists
*/
port = dp->tc_port;
if (port < 0 || port >= xsc->sc_nport) {
error = ENOENT;
goto out;
}
xpp = xsc->sc_ports + port;
regp = (struct si_reg *)xsc->sc_maddr;
}
switch (cmd) {
case TCSIDEBUG:
#ifdef SI_DEBUG
SUCHECK;
if (xpp->sp_debug)
xpp->sp_debug = 0;
else {
xpp->sp_debug = DBG_ALL;
DPRINT((xpp, DBG_IOCTL, "debug toggled %s\n",
(xpp->sp_debug&DBG_ALL)?"ON":"OFF"));
}
break;
#else
error = ENODEV;
goto out;
#endif
case TCSISDBG_LEVEL:
case TCSIGDBG_LEVEL:
#ifdef SI_DEBUG
if (cmd == TCSIGDBG_LEVEL) {
dp->tc_dbglvl = xpp->sp_debug;
} else {
SUCHECK;
xpp->sp_debug = dp->tc_dbglvl;
}
break;
#else
error = ENODEV;
goto out;
#endif
case TCSIGRXIT:
dp->tc_int = regp->rx_int_count;
break;
case TCSIRXIT:
SUCHECK;
regp->rx_int_count = dp->tc_int;
break;
case TCSIGIT:
dp->tc_int = regp->int_count;
break;
case TCSIIT:
SUCHECK;
regp->int_count = dp->tc_int;
break;
case TCSISTATE:
dp->tc_int = xpp->sp_ccb->hi_ip;
break;
/* these next three use a different structure */
case TCSI_PORT:
SUCHECK;
si_bcopy(xpp, &sps->tc_siport, sizeof(sps->tc_siport));
break;
case TCSI_CCB:
SUCHECK;
si_vbcopy(xpp->sp_ccb, &sps->tc_ccb, sizeof(sps->tc_ccb));
break;
default:
error = EINVAL;
goto out;
}
out:
splx(oldspl);
return(error); /* success */
}
/*
* Allocate a block in the file system.
*
* this takes the framework from ffs_alloc. To implement the
* actual allocation, it calls ext2_new_block, the ported version
* of the same Linux routine.
*
* we note that this is always called in connection with ext2_blkpref
*
* preallocation is done as Linux does it
*/
int
ext2_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, int size,
struct ucred *cred, daddr_t *bnp)
{
struct ext2_sb_info *fs;
daddr_t bno;
#if QUOTA
int error;
#endif
*bnp = 0;
fs = ip->i_e2fs;
#if DIAGNOSTIC
if ((u_int)size > fs->s_blocksize || blkoff(fs, size) != 0) {
kprintf("dev = %s, bsize = %lu, size = %d, fs = %s\n",
devtoname(ip->i_dev), fs->s_blocksize, size, fs->fs_fsmnt);
panic("ext2_alloc: bad size");
}
if (cred == NOCRED)
panic("ext2_alloc: missing credential");
#endif /* DIAGNOSTIC */
if (size == fs->s_blocksize && fs->s_es->s_free_blocks_count == 0)
goto nospace;
if (cred->cr_uid != 0 &&
fs->s_es->s_free_blocks_count < fs->s_es->s_r_blocks_count)
goto nospace;
#if QUOTA
if ((error = ext2_chkdq(ip, (long)btodb(size), cred, 0)) != 0)
return (error);
#endif
if (bpref >= fs->s_es->s_blocks_count)
bpref = 0;
/* call the Linux code */
#ifdef EXT2_PREALLOCATE
/* To have a preallocation hit, we must
* - have at least one block preallocated
* - and our preferred block must have that block number or one below
*/
if (ip->i_prealloc_count &&
(bpref == ip->i_prealloc_block ||
bpref + 1 == ip->i_prealloc_block))
{
bno = ip->i_prealloc_block++;
ip->i_prealloc_count--;
/* ext2_debug ("preallocation hit (%lu/%lu).\n",
++alloc_hits, ++alloc_attempts); */
/* Linux gets, clears, and releases the buffer at this
point - we don't have to that; we leave it to the caller
*/
} else {
ext2_discard_prealloc (ip);
/* ext2_debug ("preallocation miss (%lu/%lu).\n",
alloc_hits, ++alloc_attempts); */
if (S_ISREG(ip->i_mode))
bno = ext2_new_block
(ITOV(ip)->v_mount, bpref,
&ip->i_prealloc_count,
&ip->i_prealloc_block);
else
bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount,
bpref, 0, 0);
}
#else
bno = (daddr_t)ext2_new_block(ITOV(ip)->v_mount, bpref, 0, 0);
#endif
if (bno > 0) {
/* set next_alloc fields as done in block_getblk */
ip->i_next_alloc_block = lbn;
ip->i_next_alloc_goal = bno;
ip->i_blocks += btodb(size);
ip->i_flag |= IN_CHANGE | IN_UPDATE;
*bnp = bno;
return (0);
}
#if QUOTA
/*
* Restore user's disk quota because allocation failed.
*/
ext2_chkdq(ip, (long)-btodb(size), cred, FORCE);
#endif
nospace:
ext2_fserr(fs, cred->cr_uid, "file system full");
uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
return (ENOSPC);
}
int
physio(struct cdev *dev, struct uio *uio, int ioflag)
{
int i;
int error;
caddr_t sa;
u_int iolen;
struct buf *bp;
/* Keep the process UPAGES from being swapped. XXX: why ? */
PHOLD(curproc);
bp = getpbuf(NULL);
sa = bp->b_data;
error = 0;
/* XXX: sanity check */
if(dev->si_iosize_max < PAGE_SIZE) {
printf("WARNING: %s si_iosize_max=%d, using DFLTPHYS.\n",
devtoname(dev), dev->si_iosize_max);
dev->si_iosize_max = DFLTPHYS;
}
for (i = 0; i < uio->uio_iovcnt; i++) {
while (uio->uio_iov[i].iov_len) {
bp->b_flags = 0;
if (uio->uio_rw == UIO_READ) {
bp->b_iocmd = BIO_READ;
curthread->td_ru.ru_inblock++;
} else {
bp->b_iocmd = BIO_WRITE;
curthread->td_ru.ru_oublock++;
}
bp->b_iodone = bdone;
bp->b_data = uio->uio_iov[i].iov_base;
bp->b_bcount = uio->uio_iov[i].iov_len;
bp->b_offset = uio->uio_offset;
bp->b_iooffset = uio->uio_offset;
bp->b_saveaddr = sa;
/* Don't exceed drivers iosize limit */
if (bp->b_bcount > dev->si_iosize_max)
bp->b_bcount = dev->si_iosize_max;
/*
* Make sure the pbuf can map the request
* XXX: The pbuf has kvasize = MAXPHYS so a request
* XXX: larger than MAXPHYS - PAGE_SIZE must be
* XXX: page aligned or it will be fragmented.
*/
iolen = ((vm_offset_t) bp->b_data) & PAGE_MASK;
if ((bp->b_bcount + iolen) > bp->b_kvasize) {
bp->b_bcount = bp->b_kvasize;
if (iolen != 0)
bp->b_bcount -= PAGE_SIZE;
}
bp->b_bufsize = bp->b_bcount;
bp->b_blkno = btodb(bp->b_offset);
if (uio->uio_segflg == UIO_USERSPACE)
if (vmapbuf(bp) < 0) {
error = EFAULT;
goto doerror;
}
dev_strategy(dev, bp);
if (uio->uio_rw == UIO_READ)
bwait(bp, PRIBIO, "physrd");
else
bwait(bp, PRIBIO, "physwr");
if (uio->uio_segflg == UIO_USERSPACE)
vunmapbuf(bp);
iolen = bp->b_bcount - bp->b_resid;
if (iolen == 0 && !(bp->b_ioflags & BIO_ERROR))
goto doerror; /* EOF */
uio->uio_iov[i].iov_len -= iolen;
uio->uio_iov[i].iov_base =
(char *)uio->uio_iov[i].iov_base + iolen;
uio->uio_resid -= iolen;
uio->uio_offset += iolen;
if( bp->b_ioflags & BIO_ERROR) {
error = bp->b_error;
goto doerror;
}
}
}
doerror:
relpbuf(bp, NULL);
PRELE(curproc);
return (error);
}
int
physio(struct cdev *dev, struct uio *uio, int ioflag)
{
struct buf *bp;
struct cdevsw *csw;
caddr_t sa;
u_int iolen;
int error, i, mapped;
/* Keep the process UPAGES from being swapped. XXX: why ? */
PHOLD(curproc);
bp = getpbuf(NULL);
sa = bp->b_data;
error = 0;
/* XXX: sanity check */
if(dev->si_iosize_max < PAGE_SIZE) {
printf("WARNING: %s si_iosize_max=%d, using DFLTPHYS.\n",
devtoname(dev), dev->si_iosize_max);
dev->si_iosize_max = DFLTPHYS;
}
/*
* If the driver does not want I/O to be split, that means that we
* need to reject any requests that will not fit into one buffer.
*/
if (dev->si_flags & SI_NOSPLIT &&
(uio->uio_resid > dev->si_iosize_max || uio->uio_resid > MAXPHYS ||
uio->uio_iovcnt > 1)) {
/*
* Tell the user why his I/O was rejected.
*/
if (uio->uio_resid > dev->si_iosize_max)
uprintf("%s: request size=%zd > si_iosize_max=%d; "
"cannot split request\n", devtoname(dev),
uio->uio_resid, dev->si_iosize_max);
if (uio->uio_resid > MAXPHYS)
uprintf("%s: request size=%zd > MAXPHYS=%d; "
"cannot split request\n", devtoname(dev),
uio->uio_resid, MAXPHYS);
if (uio->uio_iovcnt > 1)
uprintf("%s: request vectors=%d > 1; "
"cannot split request\n", devtoname(dev),
uio->uio_iovcnt);
error = EFBIG;
goto doerror;
}
for (i = 0; i < uio->uio_iovcnt; i++) {
while (uio->uio_iov[i].iov_len) {
bp->b_flags = 0;
if (uio->uio_rw == UIO_READ) {
bp->b_iocmd = BIO_READ;
curthread->td_ru.ru_inblock++;
} else {
bp->b_iocmd = BIO_WRITE;
curthread->td_ru.ru_oublock++;
}
bp->b_iodone = bdone;
bp->b_data = uio->uio_iov[i].iov_base;
bp->b_bcount = uio->uio_iov[i].iov_len;
bp->b_offset = uio->uio_offset;
bp->b_iooffset = uio->uio_offset;
bp->b_saveaddr = sa;
/* Don't exceed drivers iosize limit */
if (bp->b_bcount > dev->si_iosize_max)
bp->b_bcount = dev->si_iosize_max;
/*
* Make sure the pbuf can map the request
* XXX: The pbuf has kvasize = MAXPHYS so a request
* XXX: larger than MAXPHYS - PAGE_SIZE must be
* XXX: page aligned or it will be fragmented.
*/
iolen = ((vm_offset_t) bp->b_data) & PAGE_MASK;
if ((bp->b_bcount + iolen) > bp->b_kvasize) {
/*
* This device does not want I/O to be split.
*/
if (dev->si_flags & SI_NOSPLIT) {
uprintf("%s: request ptr %p is not "
"on a page boundary; cannot split "
"request\n", devtoname(dev),
bp->b_data);
error = EFBIG;
goto doerror;
}
bp->b_bcount = bp->b_kvasize;
if (iolen != 0)
bp->b_bcount -= PAGE_SIZE;
}
bp->b_bufsize = bp->b_bcount;
bp->b_blkno = btodb(bp->b_offset);
csw = dev->si_devsw;
if (uio->uio_segflg == UIO_USERSPACE) {
if (dev->si_flags & SI_UNMAPPED)
mapped = 0;
else
mapped = 1;
if (vmapbuf(bp, mapped) < 0) {
error = EFAULT;
goto doerror;
}
}
dev_strategy_csw(dev, csw, bp);
if (uio->uio_rw == UIO_READ)
bwait(bp, PRIBIO, "physrd");
else
bwait(bp, PRIBIO, "physwr");
if (uio->uio_segflg == UIO_USERSPACE)
vunmapbuf(bp);
iolen = bp->b_bcount - bp->b_resid;
if (iolen == 0 && !(bp->b_ioflags & BIO_ERROR))
goto doerror; /* EOF */
uio->uio_iov[i].iov_len -= iolen;
uio->uio_iov[i].iov_base =
(char *)uio->uio_iov[i].iov_base + iolen;
uio->uio_resid -= iolen;
uio->uio_offset += iolen;
if( bp->b_ioflags & BIO_ERROR) {
error = bp->b_error;
goto doerror;
}
}
}
doerror:
relpbuf(bp, NULL);
PRELE(curproc);
return (error);
}
int
physio(struct cdev *dev, struct uio *uio, int ioflag)
{
struct cdevsw *csw;
struct buf *pbuf;
struct bio *bp;
struct vm_page **pages;
caddr_t sa;
u_int iolen, poff;
int error, i, npages, maxpages;
vm_prot_t prot;
csw = dev->si_devsw;
npages = 0;
sa = NULL;
/* check if character device is being destroyed */
if (csw == NULL)
return (ENXIO);
/* XXX: sanity check */
if(dev->si_iosize_max < PAGE_SIZE) {
printf("WARNING: %s si_iosize_max=%d, using DFLTPHYS.\n",
devtoname(dev), dev->si_iosize_max);
dev->si_iosize_max = DFLTPHYS;
}
/*
* If the driver does not want I/O to be split, that means that we
* need to reject any requests that will not fit into one buffer.
*/
if (dev->si_flags & SI_NOSPLIT &&
(uio->uio_resid > dev->si_iosize_max || uio->uio_resid > MAXPHYS ||
uio->uio_iovcnt > 1)) {
/*
* Tell the user why his I/O was rejected.
*/
if (uio->uio_resid > dev->si_iosize_max)
uprintf("%s: request size=%zd > si_iosize_max=%d; "
"cannot split request\n", devtoname(dev),
uio->uio_resid, dev->si_iosize_max);
if (uio->uio_resid > MAXPHYS)
uprintf("%s: request size=%zd > MAXPHYS=%d; "
"cannot split request\n", devtoname(dev),
uio->uio_resid, MAXPHYS);
if (uio->uio_iovcnt > 1)
uprintf("%s: request vectors=%d > 1; "
"cannot split request\n", devtoname(dev),
uio->uio_iovcnt);
return (EFBIG);
}
/*
* Keep the process UPAGES from being swapped. Processes swapped
* out while holding pbufs, used by swapper, may lead to deadlock.
*/
PHOLD(curproc);
bp = g_alloc_bio();
if (uio->uio_segflg != UIO_USERSPACE) {
pbuf = NULL;
pages = NULL;
} else if ((dev->si_flags & SI_UNMAPPED) && unmapped_buf_allowed) {
pbuf = NULL;
maxpages = btoc(MIN(uio->uio_resid, MAXPHYS)) + 1;
pages = malloc(sizeof(*pages) * maxpages, M_DEVBUF, M_WAITOK);
} else {
pbuf = uma_zalloc(pbuf_zone, M_WAITOK);
sa = pbuf->b_data;
maxpages = btoc(MAXPHYS);
pages = pbuf->b_pages;
}
prot = VM_PROT_READ;
if (uio->uio_rw == UIO_READ)
prot |= VM_PROT_WRITE; /* Less backwards than it looks */
error = 0;
for (i = 0; i < uio->uio_iovcnt; i++) {
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(curproc);
if (uio->uio_rw == UIO_READ) {
racct_add_force(curproc, RACCT_READBPS,
uio->uio_iov[i].iov_len);
racct_add_force(curproc, RACCT_READIOPS, 1);
} else {
racct_add_force(curproc, RACCT_WRITEBPS,
uio->uio_iov[i].iov_len);
racct_add_force(curproc, RACCT_WRITEIOPS, 1);
}
PROC_UNLOCK(curproc);
}
#endif /* RACCT */
while (uio->uio_iov[i].iov_len) {
g_reset_bio(bp);
if (uio->uio_rw == UIO_READ) {
bp->bio_cmd = BIO_READ;
curthread->td_ru.ru_inblock++;
} else {
bp->bio_cmd = BIO_WRITE;
curthread->td_ru.ru_oublock++;
}
bp->bio_offset = uio->uio_offset;
bp->bio_data = uio->uio_iov[i].iov_base;
bp->bio_length = uio->uio_iov[i].iov_len;
if (bp->bio_length > dev->si_iosize_max)
bp->bio_length = dev->si_iosize_max;
if (bp->bio_length > MAXPHYS)
bp->bio_length = MAXPHYS;
/*
* Make sure the pbuf can map the request.
* The pbuf has kvasize = MAXPHYS, so a request
* larger than MAXPHYS - PAGE_SIZE must be
* page aligned or it will be fragmented.
*/
poff = (vm_offset_t)bp->bio_data & PAGE_MASK;
if (pbuf && bp->bio_length + poff > pbuf->b_kvasize) {
if (dev->si_flags & SI_NOSPLIT) {
uprintf("%s: request ptr %p is not "
"on a page boundary; cannot split "
"request\n", devtoname(dev),
bp->bio_data);
error = EFBIG;
goto doerror;
}
bp->bio_length = pbuf->b_kvasize;
if (poff != 0)
bp->bio_length -= PAGE_SIZE;
}
bp->bio_bcount = bp->bio_length;
bp->bio_dev = dev;
if (pages) {
if ((npages = vm_fault_quick_hold_pages(
&curproc->p_vmspace->vm_map,
(vm_offset_t)bp->bio_data, bp->bio_length,
prot, pages, maxpages)) < 0) {
error = EFAULT;
goto doerror;
}
if (pbuf && sa) {
pmap_qenter((vm_offset_t)sa,
pages, npages);
bp->bio_data = sa + poff;
} else {
bp->bio_ma = pages;
bp->bio_ma_n = npages;
bp->bio_ma_offset = poff;
bp->bio_data = unmapped_buf;
bp->bio_flags |= BIO_UNMAPPED;
}
}
csw->d_strategy(bp);
if (uio->uio_rw == UIO_READ)
biowait(bp, "physrd");
else
biowait(bp, "physwr");
if (pages) {
if (pbuf)
pmap_qremove((vm_offset_t)sa, npages);
vm_page_unhold_pages(pages, npages);
}
iolen = bp->bio_length - bp->bio_resid;
if (iolen == 0 && !(bp->bio_flags & BIO_ERROR))
goto doerror; /* EOF */
uio->uio_iov[i].iov_len -= iolen;
uio->uio_iov[i].iov_base =
(char *)uio->uio_iov[i].iov_base + iolen;
uio->uio_resid -= iolen;
uio->uio_offset += iolen;
if (bp->bio_flags & BIO_ERROR) {
error = bp->bio_error;
goto doerror;
}
}
}
doerror:
if (pbuf)
uma_zfree(pbuf_zone, pbuf);
else if (pages)
free(pages, M_DEVBUF);
g_destroy_bio(bp);
PRELE(curproc);
return (error);
}
static
void
mbr_extended(cdev_t dev, struct disk_info *info, struct diskslices *ssp,
u_int64_t ext_offset, u_int64_t ext_size, u_int64_t base_ext_offset,
int nsectors, int ntracks, u_int64_t mbr_offset, int level)
{
struct buf *bp;
u_char *cp;
int dospart;
struct dos_partition *dp;
struct dos_partition dpcopy[NDOSPART];
u_int64_t ext_offsets[NDOSPART];
u_int64_t ext_sizes[NDOSPART];
char partname[2];
int slice;
char *sname;
struct diskslice *sp;
if (level >= 16) {
kprintf(
"%s: excessive recursion in search for slices; aborting search\n",
devtoname(dev));
return;
}
/* Read extended boot record. */
bp = geteblk((int)info->d_media_blksize);
bp->b_bio1.bio_offset = (off_t)ext_offset * info->d_media_blksize;
bp->b_bio1.bio_done = biodone_sync;
bp->b_bio1.bio_flags |= BIO_SYNC;
bp->b_bcount = info->d_media_blksize;
bp->b_cmd = BUF_CMD_READ;
bp->b_flags |= B_FAILONDIS;
dev_dstrategy(dev, &bp->b_bio1);
if (biowait(&bp->b_bio1, "mbrrd") != 0) {
diskerr(&bp->b_bio1, dev,
"reading extended partition table: error",
LOG_PRINTF, 0);
kprintf("\n");
goto done;
}
/* Weakly verify it. */
cp = bp->b_data;
if (cp[0x1FE] != 0x55 || cp[0x1FF] != 0xAA) {
sname = dsname(dev, dkunit(dev), WHOLE_DISK_SLICE, WHOLE_SLICE_PART,
partname);
if (bootverbose)
kprintf("%s: invalid extended partition table: no magic\n",
sname);
goto done;
}
/* Make a copy of the partition table to avoid alignment problems. */
memcpy(&dpcopy[0], cp + DOSPARTOFF, sizeof(dpcopy));
slice = ssp->dss_nslices;
for (dospart = 0, dp = &dpcopy[0]; dospart < NDOSPART;
dospart++, dp++) {
ext_sizes[dospart] = 0;
if (dp->dp_scyl == 0 && dp->dp_shd == 0 && dp->dp_ssect == 0
&& dp->dp_start == 0 && dp->dp_size == 0)
continue;
if (dp->dp_typ == DOSPTYP_EXTENDED ||
dp->dp_typ == DOSPTYP_EXTENDEDX) {
static char buf[32];
sname = dsname(dev, dkunit(dev), WHOLE_DISK_SLICE,
WHOLE_SLICE_PART, partname);
ksnprintf(buf, sizeof(buf), "%s", sname);
if (strlen(buf) < sizeof buf - 11)
strcat(buf, "<extended>");
check_part(buf, dp, base_ext_offset, nsectors,
ntracks, mbr_offset);
ext_offsets[dospart] = base_ext_offset + dp->dp_start;
ext_sizes[dospart] = dp->dp_size;
} else {
sname = dsname(dev, dkunit(dev), slice, WHOLE_SLICE_PART,
partname);
check_part(sname, dp, ext_offset, nsectors, ntracks,
mbr_offset);
if (slice >= MAX_SLICES) {
kprintf("%s: too many slices\n", sname);
slice++;
continue;
}
sp = &ssp->dss_slices[slice];
if (mbr_setslice(sname, info, sp, dp, ext_offset) != 0)
continue;
slice++;
}
}
ssp->dss_nslices = slice;
/* If we found any more slices, recursively find all the subslices. */
for (dospart = 0; dospart < NDOSPART; dospart++) {
if (ext_sizes[dospart] != 0) {
mbr_extended(dev, info, ssp, ext_offsets[dospart],
ext_sizes[dospart], base_ext_offset,
nsectors, ntracks, mbr_offset, ++level);
}
}
done:
bp->b_flags |= B_INVAL | B_AGE;
brelse(bp);
}
int
physio(dev_t dev, struct uio *uio, int ioflag)
{
int i;
int error;
int spl;
caddr_t sa;
off_t blockno;
u_int iolen;
struct buf *bp;
/* Keep the process UPAGES from being swapped. XXX: why ? */
PHOLD(curproc);
bp = getpbuf(NULL);
sa = bp->b_data;
error = bp->b_error = 0;
/* XXX: sanity check */
if(dev->si_iosize_max < PAGE_SIZE) {
printf("WARNING: %s si_iosize_max=%d, using DFLTPHYS.\n",
devtoname(dev), dev->si_iosize_max);
dev->si_iosize_max = DFLTPHYS;
}
for (i = 0; i < uio->uio_iovcnt; i++) {
while (uio->uio_iov[i].iov_len) {
if (uio->uio_rw == UIO_READ)
bp->b_flags = B_PHYS | B_CALL | B_READ;
else
bp->b_flags = B_PHYS | B_CALL | B_WRITE;
bp->b_dev = dev;
bp->b_iodone = physwakeup;
bp->b_data = uio->uio_iov[i].iov_base;
bp->b_bcount = uio->uio_iov[i].iov_len;
bp->b_offset = uio->uio_offset;
bp->b_saveaddr = sa;
/* Don't exceed drivers iosize limit */
if (bp->b_bcount > dev->si_iosize_max)
bp->b_bcount = dev->si_iosize_max;
/*
* Make sure the pbuf can map the request
* XXX: The pbuf has kvasize = MAXPHYS so a request
* XXX: larger than MAXPHYS - PAGE_SIZE must be
* XXX: page aligned or it will be fragmented.
*/
iolen = ((vm_offset_t) bp->b_data) & PAGE_MASK;
if ((bp->b_bcount + iolen) > bp->b_kvasize) {
bp->b_bcount = bp->b_kvasize;
if (iolen != 0)
bp->b_bcount -= PAGE_SIZE;
}
bp->b_bufsize = bp->b_bcount;
blockno = bp->b_offset >> DEV_BSHIFT;
if ((daddr_t)blockno != blockno) {
error = EINVAL; /* blockno overflow */
goto doerror;
}
bp->b_blkno = blockno;
if (uio->uio_segflg == UIO_USERSPACE) {
if (!useracc(bp->b_data, bp->b_bufsize,
bp->b_flags & B_READ ?
VM_PROT_WRITE : VM_PROT_READ)) {
error = EFAULT;
goto doerror;
}
vmapbuf(bp);
}
BUF_STRATEGY(bp, 0);
spl = splbio();
while ((bp->b_flags & B_DONE) == 0)
tsleep((caddr_t)bp, PRIBIO, "physstr", 0);
splx(spl);
if (uio->uio_segflg == UIO_USERSPACE)
vunmapbuf(bp);
iolen = bp->b_bcount - bp->b_resid;
if (iolen == 0 && !(bp->b_flags & B_ERROR))
goto doerror; /* EOF */
uio->uio_iov[i].iov_len -= iolen;
uio->uio_iov[i].iov_base += iolen;
uio->uio_resid -= iolen;
uio->uio_offset += iolen;
if( bp->b_flags & B_ERROR) {
error = bp->b_error;
goto doerror;
}
}
}
doerror:
relpbuf(bp, NULL);
PRELE(curproc);
return (error);
}
/*
* spec_getpages() - get pages associated with device vnode.
*
* Note that spec_read and spec_write do not use the buffer cache, so we
* must fully implement getpages here.
*/
static int
devfs_spec_getpages(struct vop_getpages_args *ap)
{
vm_offset_t kva;
int error;
int i, pcount, size;
struct buf *bp;
vm_page_t m;
vm_ooffset_t offset;
int toff, nextoff, nread;
struct vnode *vp = ap->a_vp;
int blksiz;
int gotreqpage;
error = 0;
pcount = round_page(ap->a_count) / PAGE_SIZE;
/*
* Calculate the offset of the transfer and do sanity check.
*/
offset = IDX_TO_OFF(ap->a_m[0]->pindex) + ap->a_offset;
/*
* Round up physical size for real devices. We cannot round using
* v_mount's block size data because v_mount has nothing to do with
* the device. i.e. it's usually '/dev'. We need the physical block
* size for the device itself.
*
* We can't use v_rdev->si_mountpoint because it only exists when the
* block device is mounted. However, we can use v_rdev.
*/
if (vn_isdisk(vp, NULL))
blksiz = vp->v_rdev->si_bsize_phys;
else
blksiz = DEV_BSIZE;
size = (ap->a_count + blksiz - 1) & ~(blksiz - 1);
bp = getpbuf_kva(NULL);
kva = (vm_offset_t)bp->b_data;
/*
* Map the pages to be read into the kva.
*/
pmap_qenter(kva, ap->a_m, pcount);
/* Build a minimal buffer header. */
bp->b_cmd = BUF_CMD_READ;
bp->b_bcount = size;
bp->b_resid = 0;
bsetrunningbufspace(bp, size);
bp->b_bio1.bio_offset = offset;
bp->b_bio1.bio_done = devfs_spec_getpages_iodone;
mycpu->gd_cnt.v_vnodein++;
mycpu->gd_cnt.v_vnodepgsin += pcount;
/* Do the input. */
vn_strategy(ap->a_vp, &bp->b_bio1);
crit_enter();
/* We definitely need to be at splbio here. */
while (bp->b_cmd != BUF_CMD_DONE)
tsleep(bp, 0, "spread", 0);
crit_exit();
if (bp->b_flags & B_ERROR) {
if (bp->b_error)
error = bp->b_error;
else
error = EIO;
}
/*
* If EOF is encountered we must zero-extend the result in order
* to ensure that the page does not contain garabge. When no
* error occurs, an early EOF is indicated if b_bcount got truncated.
* b_resid is relative to b_bcount and should be 0, but some devices
* might indicate an EOF with b_resid instead of truncating b_bcount.
*/
nread = bp->b_bcount - bp->b_resid;
if (nread < ap->a_count)
bzero((caddr_t)kva + nread, ap->a_count - nread);
pmap_qremove(kva, pcount);
gotreqpage = 0;
for (i = 0, toff = 0; i < pcount; i++, toff = nextoff) {
nextoff = toff + PAGE_SIZE;
m = ap->a_m[i];
m->flags &= ~PG_ZERO;
/*
* NOTE: vm_page_undirty/clear_dirty etc do not clear the
* pmap modified bit. pmap modified bit should have
* already been cleared.
*/
if (nextoff <= nread) {
m->valid = VM_PAGE_BITS_ALL;
vm_page_undirty(m);
} else if (toff < nread) {
/*
* Since this is a VM request, we have to supply the
* unaligned offset to allow vm_page_set_valid()
* to zero sub-DEV_BSIZE'd portions of the page.
*/
vm_page_set_valid(m, 0, nread - toff);
vm_page_clear_dirty_end_nonincl(m, 0, nread - toff);
} else {
m->valid = 0;
vm_page_undirty(m);
}
if (i != ap->a_reqpage) {
/*
* Just in case someone was asking for this page we
* now tell them that it is ok to use.
*/
if (!error || (m->valid == VM_PAGE_BITS_ALL)) {
if (m->valid) {
if (m->flags & PG_REFERENCED) {
vm_page_activate(m);
} else {
vm_page_deactivate(m);
}
vm_page_wakeup(m);
} else {
vm_page_free(m);
}
} else {
vm_page_free(m);
}
} else if (m->valid) {
gotreqpage = 1;
/*
* Since this is a VM request, we need to make the
* entire page presentable by zeroing invalid sections.
*/
if (m->valid != VM_PAGE_BITS_ALL)
vm_page_zero_invalid(m, FALSE);
}
}
if (!gotreqpage) {
m = ap->a_m[ap->a_reqpage];
devfs_debug(DEVFS_DEBUG_WARNING,
"spec_getpages:(%s) I/O read failure: (error=%d) bp %p vp %p\n",
devtoname(vp->v_rdev), error, bp, bp->b_vp);
devfs_debug(DEVFS_DEBUG_WARNING,
" size: %d, resid: %d, a_count: %d, valid: 0x%x\n",
size, bp->b_resid, ap->a_count, m->valid);
devfs_debug(DEVFS_DEBUG_WARNING,
" nread: %d, reqpage: %d, pindex: %lu, pcount: %d\n",
nread, ap->a_reqpage, (u_long)m->pindex, pcount);
/*
* Free the buffer header back to the swap buffer pool.
*/
relpbuf(bp, NULL);
return VM_PAGER_ERROR;
}
/*
* Free the buffer header back to the swap buffer pool.
*/
relpbuf(bp, NULL);
if (DEVFS_NODE(ap->a_vp))
nanotime(&DEVFS_NODE(ap->a_vp)->mtime);
return VM_PAGER_OK;
}
