Differential D23010 Diff 69736 usr.sbin/bhyve/block_local.c

Changeset View

Standalone View

usr.sbin/bhyve/block_local.c

This file was copied from usr.sbin/bhyve/block_if.c.

/*-		/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD		* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*		*
* Copyright (c) 2013 Peter Grehan <grehan@freebsd.org>		* Copyright (c) 2020 Willem Jan Withagen <wjw@digiware.nl>
* All rights reserved.		* All rights reserved.
*		*
* Redistribution and use in source and binary forms, with or without		* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions		* modification, are permitted provided that the following conditions
* are met:		* are met:
* 1. Redistributions of source code must retain the above copyright		* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.		* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright		* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the		* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.		* documentation and/or other materials provided with the distribution.
*		*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND		* THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE		* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE		* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE		* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL		* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS		* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)		* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT		* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY		* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF		* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.		* SUCH DAMAGE.
*		*
* $FreeBSD: head/usr.sbin/bhyve/block_if.c 356523 2020-01-08 22:55:22Z vmaffione $		* $FreeBSD$
*/		*/

#include <sys/cdefs.h>		#include <sys/cdefs.h>
__FBSDID("$FreeBSD: head/usr.sbin/bhyve/block_if.c 356523 2020-01-08 22:55:22Z vmaffione $");		__FBSDID("$FreeBSD$");

#include <sys/param.h>		#include <sys/param.h>
#ifndef WITHOUT_CAPSICUM		#ifndef WITHOUT_CAPSICUM
#include <sys/capsicum.h>		#include <sys/capsicum.h>
#endif		#endif
#include <sys/queue.h>
#include <sys/errno.h>		#include <sys/errno.h>
#include <sys/stat.h>		#include <sys/stat.h>
#include <sys/ioctl.h>		#include <sys/ioctl.h>
#include <sys/disk.h>		#include <sys/disk.h>

#include <assert.h>		#include <assert.h>
#ifndef WITHOUT_CAPSICUM		#ifndef WITHOUT_CAPSICUM
#include <capsicum_helpers.h>		#include <capsicum_helpers.h>
Show All 11 Lines

#include <machine/atomic.h>		#include <machine/atomic.h>

#include "bhyverun.h"		#include "bhyverun.h"
#include "debug.h"		#include "debug.h"
#include "mevent.h"		#include "mevent.h"
#include "block_if.h"		#include "block_if.h"

#define BLOCKIF_SIG 0xb109b109		#include <sys/linker_set.h>

#define BLOCKIF_NUMTHR 8		static pthread_once_t blocklocal_once = PTHREAD_ONCE_INIT;
#define BLOCKIF_MAXREQ (BLOCKIF_RING_MAX + BLOCKIF_NUMTHR)

enum blockop {		struct blocklocal_sig_elem {
BOP_READ,
BOP_WRITE,
BOP_FLUSH,
BOP_DELETE
};

enum blockstat {
BST_FREE,
BST_BLOCK,
BST_PEND,
BST_BUSY,
BST_DONE
};

struct blockif_elem {
TAILQ_ENTRY(blockif_elem) be_link;
struct blockif_req *be_req;
enum blockop be_op;
enum blockstat be_status;
pthread_t be_tid;
off_t be_block;
};

struct blockif_ctxt {
int bc_magic;
int bc_fd;
int bc_ischr;
int bc_isgeom;
int bc_candelete;
int bc_rdonly;
off_t bc_size;
int bc_sectsz;
int bc_psectsz;
int bc_psectoff;
int bc_closing;
pthread_t bc_btid[BLOCKIF_NUMTHR];
pthread_mutex_t bc_mtx;
pthread_cond_t bc_cond;

/* Request elements and free/pending/busy queues */
TAILQ_HEAD(, blockif_elem) bc_freeq;
TAILQ_HEAD(, blockif_elem) bc_pendq;
TAILQ_HEAD(, blockif_elem) bc_busyq;
struct blockif_elem bc_reqs[BLOCKIF_MAXREQ];
};

static pthread_once_t blockif_once = PTHREAD_ONCE_INIT;

struct blockif_sig_elem {
pthread_mutex_t bse_mtx;		pthread_mutex_t bse_mtx;
pthread_cond_t bse_cond;		pthread_cond_t bse_cond;
int bse_pending;		int bse_pending;
struct blockif_sig_elem *bse_next;		struct blocklocal_sig_elem *bse_next;
};		};

static struct blockif_sig_elem *blockif_bse_head;		static struct blocklocal_sig_elem *blocklocal_bse_head;

static int		static int
blockif_enqueue(struct blockif_ctxt bc, struct blockif_req breq,		blocklocal_enqueue(struct blockif_ctxt bc, struct blockif_req breq,
enum blockop op)		enum blockop op)
{		{
struct blockif_elem be, tbe;		struct blockif_elem be, tbe;
off_t off;		off_t off;
int i;		int i;

be = TAILQ_FIRST(&bc->bc_freeq);		be = TAILQ_FIRST(&bc->bc_freeq);
assert(be != NULL);		assert(be != NULL);
Show All 27 Lines	if (tbe == NULL)
be->be_status = BST_PEND;		be->be_status = BST_PEND;
else		else
be->be_status = BST_BLOCK;		be->be_status = BST_BLOCK;
TAILQ_INSERT_TAIL(&bc->bc_pendq, be, be_link);		TAILQ_INSERT_TAIL(&bc->bc_pendq, be, be_link);
return (be->be_status == BST_PEND);		return (be->be_status == BST_PEND);
}		}

static int		static int
blockif_dequeue(struct blockif_ctxt bc, pthread_t t, struct blockif_elem *bep)		blocklocal_dequeue(struct blockif_ctxt bc, pthread_t t, struct blockif_elem *bep)
{		{
struct blockif_elem *be;		struct blockif_elem *be;

TAILQ_FOREACH(be, &bc->bc_pendq, be_link) {		TAILQ_FOREACH(be, &bc->bc_pendq, be_link) {
if (be->be_status == BST_PEND)		if (be->be_status == BST_PEND)
break;		break;
assert(be->be_status == BST_BLOCK);		assert(be->be_status == BST_BLOCK);
}		}
if (be == NULL)		if (be == NULL)
return (0);		return (0);
TAILQ_REMOVE(&bc->bc_pendq, be, be_link);		TAILQ_REMOVE(&bc->bc_pendq, be, be_link);
be->be_status = BST_BUSY;		be->be_status = BST_BUSY;
be->be_tid = t;		be->be_tid = t;
TAILQ_INSERT_TAIL(&bc->bc_busyq, be, be_link);		TAILQ_INSERT_TAIL(&bc->bc_busyq, be, be_link);
*bep = be;		*bep = be;
return (1);		return (1);
}		}

static void		static void
blockif_complete(struct blockif_ctxt bc, struct blockif_elem be)		blocklocal_complete(struct blockif_ctxt bc, struct blockif_elem be)
{		{
struct blockif_elem *tbe;		struct blockif_elem *tbe;

if (be->be_status == BST_DONE \|\| be->be_status == BST_BUSY)		if (be->be_status == BST_DONE \|\| be->be_status == BST_BUSY)
TAILQ_REMOVE(&bc->bc_busyq, be, be_link);		TAILQ_REMOVE(&bc->bc_busyq, be, be_link);
else		else
TAILQ_REMOVE(&bc->bc_pendq, be, be_link);		TAILQ_REMOVE(&bc->bc_pendq, be, be_link);
TAILQ_FOREACH(tbe, &bc->bc_pendq, be_link) {		TAILQ_FOREACH(tbe, &bc->bc_pendq, be_link) {
if (tbe->be_req->br_offset == be->be_block)		if (tbe->be_req->br_offset == be->be_block)
tbe->be_status = BST_PEND;		tbe->be_status = BST_PEND;
}		}
be->be_tid = 0;		be->be_tid = 0;
be->be_status = BST_FREE;		be->be_status = BST_FREE;
be->be_req = NULL;		be->be_req = NULL;
TAILQ_INSERT_TAIL(&bc->bc_freeq, be, be_link);		TAILQ_INSERT_TAIL(&bc->bc_freeq, be, be_link);
}		}

static void		static void
blockif_proc(struct blockif_ctxt bc, struct blockif_elem be, uint8_t *buf)		blocklocal_proc(struct blockif_ctxt bc, struct blockif_elem be, uint8_t *buf)
{		{
struct blockif_req *br;		struct blockif_req *br;
off_t arg[2];		off_t arg[2];
ssize_t clen, len, off, boff, voff;		ssize_t clen, len, off, boff, voff;
int i, err;		int i, err;

br = be->be_req;		br = be->be_req;
if (br->br_iovcnt <= 1)		if (br->br_iovcnt <= 1)
buf = NULL;		buf = NULL;
err = 0;		err = 0;
switch (be->be_op) {		switch (be->be_op) {
case BOP_READ:		case BOP_READ:
if (buf == NULL) {		if (buf == NULL) {
if ((len = preadv(bc->bc_fd, br->br_iov, br->br_iovcnt,		if ((len = preadv((int)bc->bc_desc, br->br_iov, br->br_iovcnt,
br->br_offset)) < 0)		br->br_offset)) < 0)
err = errno;		err = errno;
else		else
br->br_resid -= len;		br->br_resid -= len;
break;		break;
}		}
i = 0;		i = 0;
off = voff = 0;		off = voff = 0;
while (br->br_resid > 0) {		while (br->br_resid > 0) {
len = MIN(br->br_resid, MAXPHYS);		len = MIN(br->br_resid, MAXPHYS);
if (pread(bc->bc_fd, buf, len, br->br_offset +		if (pread((int)bc->bc_desc, buf, len, br->br_offset +
off) < 0) {		off) < 0) {
err = errno;		err = errno;
break;		break;
}		}
boff = 0;		boff = 0;
do {		do {
clen = MIN(len - boff, br->br_iov[i].iov_len -		clen = MIN(len - boff, br->br_iov[i].iov_len -
voff);		voff);
Show All 12 Lines	case BOP_READ:
}		}
break;		break;
case BOP_WRITE:		case BOP_WRITE:
if (bc->bc_rdonly) {		if (bc->bc_rdonly) {
err = EROFS;		err = EROFS;
break;		break;
}		}
if (buf == NULL) {		if (buf == NULL) {
if ((len = pwritev(bc->bc_fd, br->br_iov, br->br_iovcnt,		if ((len = pwritev((int)bc->bc_desc, br->br_iov, br->br_iovcnt,
br->br_offset)) < 0)		br->br_offset)) < 0)
err = errno;		err = errno;
else		else
br->br_resid -= len;		br->br_resid -= len;
break;		break;
}		}
i = 0;		i = 0;
off = voff = 0;		off = voff = 0;
while (br->br_resid > 0) {		while (br->br_resid > 0) {
len = MIN(br->br_resid, MAXPHYS);		len = MIN(br->br_resid, MAXPHYS);
boff = 0;		boff = 0;
do {		do {
clen = MIN(len - boff, br->br_iov[i].iov_len -		clen = MIN(len - boff, br->br_iov[i].iov_len -
voff);		voff);
memcpy(buf + boff,		memcpy(buf + boff,
br->br_iov[i].iov_base + voff, clen);		br->br_iov[i].iov_base + voff, clen);
if (clen < br->br_iov[i].iov_len - voff)		if (clen < br->br_iov[i].iov_len - voff)
voff += clen;		voff += clen;
else {		else {
i++;		i++;
voff = 0;		voff = 0;
}		}
boff += clen;		boff += clen;
} while (boff < len);		} while (boff < len);
if (pwrite(bc->bc_fd, buf, len, br->br_offset +		if (pwrite((int)bc->bc_desc, buf, len, br->br_offset +
off) < 0) {		off) < 0) {
err = errno;		err = errno;
break;		break;
}		}
off += len;		off += len;
br->br_resid -= len;		br->br_resid -= len;
}		}
break;		break;
case BOP_FLUSH:		case BOP_FLUSH:
if (bc->bc_ischr) {		if (bc->bc_ischr) {
if (ioctl(bc->bc_fd, DIOCGFLUSH))		if (ioctl((int)bc->bc_desc, DIOCGFLUSH))
err = errno;		err = errno;
} else if (fsync(bc->bc_fd))		} else if (fsync((int)bc->bc_desc))
err = errno;		err = errno;
break;		break;
case BOP_DELETE:		case BOP_DELETE:
if (!bc->bc_candelete)		if (!bc->bc_candelete)
err = EOPNOTSUPP;		err = EOPNOTSUPP;
else if (bc->bc_rdonly)		else if (bc->bc_rdonly)
err = EROFS;		err = EROFS;
else if (bc->bc_ischr) {		else if (bc->bc_ischr) {
arg[0] = br->br_offset;		arg[0] = br->br_offset;
arg[1] = br->br_resid;		arg[1] = br->br_resid;
if (ioctl(bc->bc_fd, DIOCGDELETE, arg))		if (ioctl((int)bc->bc_desc, DIOCGDELETE, arg))
err = errno;		err = errno;
else		else
br->br_resid = 0;		br->br_resid = 0;
} else		} else
err = EOPNOTSUPP;		err = EOPNOTSUPP;
break;		break;
default:		default:
err = EINVAL;		err = EINVAL;
break;		break;
}		}

be->be_status = BST_DONE;		be->be_status = BST_DONE;

(*br->br_callback)(br, err);		(*br->br_callback)(br, err);
}		}

static void *		static void *
blockif_thr(void *arg)		blocklocal_thr(void *arg)
{		{
struct blockif_ctxt *bc;		struct blockif_ctxt *bc;
struct blockif_elem *be;		struct blockif_elem *be;
pthread_t t;		pthread_t t;
uint8_t *buf;		uint8_t *buf;

bc = arg;		bc = arg;
if (bc->bc_isgeom)		if (bc->bc_isgeom)
buf = malloc(MAXPHYS);		buf = malloc(MAXPHYS);
else		else
buf = NULL;		buf = NULL;
t = pthread_self();		t = pthread_self();

pthread_mutex_lock(&bc->bc_mtx);		pthread_mutex_lock(&bc->bc_mtx);
for (;;) {		for (;;) {
while (blockif_dequeue(bc, t, &be)) {		while (blocklocal_dequeue(bc, t, &be)) {
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);
blockif_proc(bc, be, buf);		blocklocal_proc(bc, be, buf);
pthread_mutex_lock(&bc->bc_mtx);		pthread_mutex_lock(&bc->bc_mtx);
blockif_complete(bc, be);		blocklocal_complete(bc, be);
}		}
/* Check ctxt status here to see if exit requested */		/* Check ctxt status here to see if exit requested */
if (bc->bc_closing)		if (bc->bc_closing)
break;		break;
pthread_cond_wait(&bc->bc_cond, &bc->bc_mtx);		pthread_cond_wait(&bc->bc_cond, &bc->bc_mtx);
}		}
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);

if (buf)		if (buf)
free(buf);		free(buf);
pthread_exit(NULL);		pthread_exit(NULL);
return (NULL);		return (NULL);
}		}

static void		static void
blockif_sigcont_handler(int signal, enum ev_type type, void *arg)		blocklocal_sigcont_handler(int signal, enum ev_type type, void *arg)
{		{
struct blockif_sig_elem *bse;		struct blocklocal_sig_elem *bse;

for (;;) {		for (;;) {
/*		/*
* Process the entire list even if not intended for		* Process the entire list even if not intended for
* this thread.		* this thread.
*/		*/
do {		do {
bse = blockif_bse_head;		bse = blocklocal_bse_head;
if (bse == NULL)		if (bse == NULL)
return;		return;
} while (!atomic_cmpset_ptr((uintptr_t *)&blockif_bse_head,		} while (!atomic_cmpset_ptr((uintptr_t *)&blocklocal_bse_head,
(uintptr_t)bse,		(uintptr_t)bse,
(uintptr_t)bse->bse_next));		(uintptr_t)bse->bse_next));

pthread_mutex_lock(&bse->bse_mtx);		pthread_mutex_lock(&bse->bse_mtx);
bse->bse_pending = 0;		bse->bse_pending = 0;
pthread_cond_signal(&bse->bse_cond);		pthread_cond_signal(&bse->bse_cond);
pthread_mutex_unlock(&bse->bse_mtx);		pthread_mutex_unlock(&bse->bse_mtx);
}		}
}		}

static void		void
blockif_init(void)		blocklocal_init(void)
{		{
mevent_add(SIGCONT, EVF_SIGNAL, blockif_sigcont_handler, NULL);		mevent_add(SIGCONT, EVF_SIGNAL, blocklocal_sigcont_handler, NULL);
(void) signal(SIGCONT, SIG_IGN);		(void) signal(SIGCONT, SIG_IGN);
}		}

struct blockif_ctxt *		void
blockif_open(const char optstr, const char ident)		blocklocal_cleanup(struct blockif_ctxt *bc)
		{ /* empty block
		* currently no cleanup required.
		*/
		}

		static struct blockif_ctxt *
		blocklocal_open(const char optstr, const char ident)
{		{
char tname[MAXCOMLEN + 1];		char tname[MAXCOMLEN + 1];
char name[MAXPATHLEN];		char name[MAXPATHLEN];
char nopt, xopts, *cp;		char nopt, xopts, *cp;
struct blockif_ctxt *bc;		struct blockif_ctxt *bc;
struct stat sbuf;		struct stat sbuf;
struct diocgattr_arg arg;		struct diocgattr_arg arg;
off_t size, psectsz, psectoff;		off_t size, psectsz, psectoff;
int extra, fd, i, sectsz;		int extra, fd, i, sectsz;
int nocache, sync, ro, candelete, geom, ssopt, pssopt;		int nocache, sync, ro, candelete, geom, ssopt, pssopt;
#ifndef WITHOUT_CAPSICUM		#ifndef WITHOUT_CAPSICUM
cap_rights_t rights;		cap_rights_t rights;
cap_ioctl_t cmds[] = { DIOCGFLUSH, DIOCGDELETE };		cap_ioctl_t cmds[] = { DIOCGFLUSH, DIOCGDELETE };
#endif		#endif

pthread_once(&blockif_once, blockif_init);		pthread_once(&blocklocal_once, blocklocal_init);

fd = -1;		fd = -1;
ssopt = 0;		ssopt = 0;
nocache = 0;		nocache = 0;
sync = 0;		sync = 0;
ro = 0;		ro = 0;

/*		/*
Show All 34 Lines	if (fd < 0 && !ro) {
ro = 1;		ro = 1;
}		}

if (fd < 0) {		if (fd < 0) {
warn("Could not open backing file: %s", nopt);		warn("Could not open backing file: %s", nopt);
goto err;		goto err;
}		}

if (fstat(fd, &sbuf) < 0) {		if (fstat(fd, &sbuf) < 0) {
warn("Could not stat backing file %s", nopt);		warn("Could not stat backing file %s", nopt);
goto err;		goto err;
}		}

#ifndef WITHOUT_CAPSICUM		#ifndef WITHOUT_CAPSICUM
cap_rights_init(&rights, CAP_FSYNC, CAP_IOCTL, CAP_READ, CAP_SEEK,		cap_rights_init(&rights, CAP_FSYNC, CAP_IOCTL, CAP_READ, CAP_SEEK,
CAP_WRITE);		CAP_WRITE);
if (ro)		if (ro)
cap_rights_clear(&rights, CAP_FSYNC, CAP_WRITE);		cap_rights_clear(&rights, CAP_FSYNC, CAP_WRITE);

if (caph_rights_limit(fd, &rights) == -1)		if (caph_rights_limit(fd, &rights) == -1)
errx(EX_OSERR, "Unable to apply rights for sandbox");		errx(EX_OSERR, "Unable to apply rights for sandbox");
#endif		#endif

/*		/*
* Deal with raw devices		* Deal with raw devices
*/		*/
size = sbuf.st_size;		size = sbuf.st_size;
sectsz = DEV_BSIZE;		sectsz = DEV_BSIZE;
psectsz = psectoff = 0;		psectsz = psectoff = 0;
candelete = geom = 0;		candelete = geom = 0;
if (S_ISCHR(sbuf.st_mode)) {		if (S_ISCHR(sbuf.st_mode)) {
if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 \|\|		if (ioctl(fd, DIOCGMEDIASIZE, &size) < 0 \|\|
ioctl(fd, DIOCGSECTORSIZE, &sectsz)) {		ioctl(fd, DIOCGSECTORSIZE, &sectsz)) {
perror("Could not fetch dev blk/sector size");		perror("Could not fetch dev blk/sector size");
goto err;		goto err;
▲ Show 20 Lines • Show All 47 Lines • ▼ Show 20 Lines	#endif

bc = calloc(1, sizeof(struct blockif_ctxt));		bc = calloc(1, sizeof(struct blockif_ctxt));
if (bc == NULL) {		if (bc == NULL) {
perror("calloc");		perror("calloc");
goto err;		goto err;
}		}

bc->bc_magic = BLOCKIF_SIG;		bc->bc_magic = BLOCKIF_SIG;
bc->bc_fd = fd;		bc->bc_desc = fd;
bc->bc_ischr = S_ISCHR(sbuf.st_mode);		bc->bc_ischr = S_ISCHR(sbuf.st_mode);
bc->bc_isgeom = geom;		bc->bc_isgeom = geom;
bc->bc_candelete = candelete;		bc->bc_candelete = candelete;
bc->bc_rdonly = ro;		bc->bc_rdonly = ro;
bc->bc_size = size;		bc->bc_size = size;
bc->bc_sectsz = sectsz;		bc->bc_sectsz = sectsz;
bc->bc_psectsz = psectsz;		bc->bc_psectsz = psectsz;
bc->bc_psectoff = psectoff;		bc->bc_psectoff = psectoff;
pthread_mutex_init(&bc->bc_mtx, NULL);		pthread_mutex_init(&bc->bc_mtx, NULL);
pthread_cond_init(&bc->bc_cond, NULL);		pthread_cond_init(&bc->bc_cond, NULL);
TAILQ_INIT(&bc->bc_freeq);		TAILQ_INIT(&bc->bc_freeq);
TAILQ_INIT(&bc->bc_pendq);		TAILQ_INIT(&bc->bc_pendq);
TAILQ_INIT(&bc->bc_busyq);		TAILQ_INIT(&bc->bc_busyq);
for (i = 0; i < BLOCKIF_MAXREQ; i++) {		for (i = 0; i < BLOCKIF_MAXREQ; i++) {
bc->bc_reqs[i].be_status = BST_FREE;		bc->bc_reqs[i].be_status = BST_FREE;
TAILQ_INSERT_HEAD(&bc->bc_freeq, &bc->bc_reqs[i], be_link);		TAILQ_INSERT_HEAD(&bc->bc_freeq, &bc->bc_reqs[i], be_link);
}		}

for (i = 0; i < BLOCKIF_NUMTHR; i++) {		for (i = 0; i < BLOCKIF_NUMTHR; i++) {
pthread_create(&bc->bc_btid[i], NULL, blockif_thr, bc);		pthread_create(&bc->bc_btid[i], NULL, blocklocal_thr, bc);
snprintf(tname, sizeof(tname), "blk-%s-%d", ident, i);		snprintf(tname, sizeof(tname), "blk-%s-%d", ident, i);
pthread_set_name_np(bc->bc_btid[i], tname);		pthread_set_name_np(bc->bc_btid[i], tname);
}		}

return (bc);		return (bc);
err:		err:
if (fd >= 0)		if (fd >= 0)
close(fd);		close(fd);
Show All 10 Lines	blockif_request(struct blockif_ctxt bc, struct blockif_req breq,
err = 0;		err = 0;

pthread_mutex_lock(&bc->bc_mtx);		pthread_mutex_lock(&bc->bc_mtx);
if (!TAILQ_EMPTY(&bc->bc_freeq)) {		if (!TAILQ_EMPTY(&bc->bc_freeq)) {
/*		/*
* Enqueue and inform the block i/o thread		* Enqueue and inform the block i/o thread
* that there is work available		* that there is work available
*/		*/
if (blockif_enqueue(bc, breq, op))		if (blocklocal_enqueue(bc, breq, op))
pthread_cond_signal(&bc->bc_cond);		pthread_cond_signal(&bc->bc_cond);
} else {		} else {
/*		/*
* Callers are not allowed to enqueue more than		* Callers are not allowed to enqueue more than
* the specified blockif queue limit. Return an		* the specified blockif queue limit. Return an
* error to indicate that the queue length has been		* error to indicate that the queue length has been
* exceeded.		* exceeded.
*/		*/
err = E2BIG;		err = E2BIG;
}		}
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);

return (err);		return (err);
}		}

int		static int
blockif_read(struct blockif_ctxt bc, struct blockif_req breq)		blocklocal_read(struct blockif_ctxt bc, struct blockif_req breq)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (blockif_request(bc, breq, BOP_READ));		return (blockif_request(bc, breq, BOP_READ));
}		}

int		static int
blockif_write(struct blockif_ctxt bc, struct blockif_req breq)		blocklocal_write(struct blockif_ctxt bc, struct blockif_req breq)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (blockif_request(bc, breq, BOP_WRITE));		return (blockif_request(bc, breq, BOP_WRITE));
}		}

int		static int
blockif_flush(struct blockif_ctxt bc, struct blockif_req breq)		blocklocal_flush(struct blockif_ctxt bc, struct blockif_req breq)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (blockif_request(bc, breq, BOP_FLUSH));		return (blockif_request(bc, breq, BOP_FLUSH));
}		}

int		static int
blockif_delete(struct blockif_ctxt bc, struct blockif_req breq)		blocklocal_delete(struct blockif_ctxt bc, struct blockif_req breq)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (blockif_request(bc, breq, BOP_DELETE));		return (blockif_request(bc, breq, BOP_DELETE));
}		}

int		static int
blockif_cancel(struct blockif_ctxt bc, struct blockif_req breq)		blocklocal_cancel(struct blockif_ctxt bc, struct blockif_req breq)
{		{
struct blockif_elem *be;		struct blockif_elem *be;

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);

pthread_mutex_lock(&bc->bc_mtx);		pthread_mutex_lock(&bc->bc_mtx);
/*		/*
* Check pending requests.		* Check pending requests.
*/		*/
TAILQ_FOREACH(be, &bc->bc_pendq, be_link) {		TAILQ_FOREACH(be, &bc->bc_pendq, be_link) {
if (be->be_req == breq)		if (be->be_req == breq)
break;		break;
}		}
if (be != NULL) {		if (be != NULL) {
/*		/*
* Found it.		* Found it.
*/		*/
blockif_complete(bc, be);		blocklocal_complete(bc, be);
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);

return (0);		return (0);
}		}

/*		/*
* Check in-flight requests.		* Check in-flight requests.
*/		*/
Show All 9 Lines	if (be == NULL) {
return (EINVAL);		return (EINVAL);
}		}

/*		/*
* Interrupt the processing thread to force it return		* Interrupt the processing thread to force it return
* prematurely via it's normal callback path.		* prematurely via it's normal callback path.
*/		*/
while (be->be_status == BST_BUSY) {		while (be->be_status == BST_BUSY) {
struct blockif_sig_elem bse, *old_head;		struct blocklocal_sig_elem bse, *old_head;

pthread_mutex_init(&bse.bse_mtx, NULL);		pthread_mutex_init(&bse.bse_mtx, NULL);
pthread_cond_init(&bse.bse_cond, NULL);		pthread_cond_init(&bse.bse_cond, NULL);

bse.bse_pending = 1;		bse.bse_pending = 1;

do {		do {
old_head = blockif_bse_head;		old_head = blocklocal_bse_head;
bse.bse_next = old_head;		bse.bse_next = old_head;
} while (!atomic_cmpset_ptr((uintptr_t *)&blockif_bse_head,		} while (!atomic_cmpset_ptr((uintptr_t *)&blocklocal_bse_head,
(uintptr_t)old_head,		(uintptr_t)old_head,
(uintptr_t)&bse));		(uintptr_t)&bse));

pthread_kill(be->be_tid, SIGCONT);		pthread_kill(be->be_tid, SIGCONT);

pthread_mutex_lock(&bse.bse_mtx);		pthread_mutex_lock(&bse.bse_mtx);
while (bse.bse_pending)		while (bse.bse_pending)
pthread_cond_wait(&bse.bse_cond, &bse.bse_mtx);		pthread_cond_wait(&bse.bse_cond, &bse.bse_mtx);
pthread_mutex_unlock(&bse.bse_mtx);		pthread_mutex_unlock(&bse.bse_mtx);
}		}

pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);

/*		/*
* The processing thread has been interrupted. Since it's not		* The processing thread has been interrupted. Since it's not
* clear if the callback has been invoked yet, return EBUSY.		* clear if the callback has been invoked yet, return EBUSY.
*/		*/
return (EBUSY);		return (EBUSY);
}		}

int		static int
blockif_close(struct blockif_ctxt *bc)		blocklocal_close(struct blockif_ctxt *bc)
{		{
void *jval;		void *jval;
int i;		int i;

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);

/*		/*
* Stop the block i/o thread		* Stop the block i/o thread
*/		*/
pthread_mutex_lock(&bc->bc_mtx);		pthread_mutex_lock(&bc->bc_mtx);
bc->bc_closing = 1;		bc->bc_closing = 1;
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);
pthread_cond_broadcast(&bc->bc_cond);		pthread_cond_broadcast(&bc->bc_cond);
for (i = 0; i < BLOCKIF_NUMTHR; i++)		for (i = 0; i < BLOCKIF_NUMTHR; i++)
pthread_join(bc->bc_btid[i], &jval);		pthread_join(bc->bc_btid[i], &jval);

/* XXX Cancel queued i/o's ??? */		/* XXX Cancel queued i/o's ??? */

/*		/*
* Release resources		* Release resources
*/		*/
bc->bc_magic = 0;		bc->bc_magic = 0;
close(bc->bc_fd);		close((int)bc->bc_desc);
free(bc);		free(bc);

return (0);		return (0);
}		}

/*		/*
* Return virtual C/H/S values for a given block. Use the algorithm		* Return virtual C/H/S values for a given block. Use the algorithm
* outlined in the VHD specification to calculate values.		* outlined in the VHD specification to calculate values.
*/		*/
void		static void
blockif_chs(struct blockif_ctxt bc, uint16_t c, uint8_t h, uint8_t s)		blocklocal_chs(struct blockif_ctxt bc, uint16_t c, uint8_t h, uint8_t s)
{		{
off_t sectors; /* total sectors of the block dev */		off_t sectors; /* total sectors of the block dev */
off_t hcyl; /* cylinders times heads */		off_t hcyl; /* cylinders times heads */
uint16_t secpt; /* sectors per track */		uint16_t secpt; /* sectors per track */
uint8_t heads;		uint8_t heads;

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);

Show All 30 Lines	blocklocal_chs(struct blockif_ctxt bc, uint16_t c, uint8_t h, uint8_t s)
*c = hcyl / heads;		*c = hcyl / heads;
*h = heads;		*h = heads;
*s = secpt;		*s = secpt;
}		}

/*		/*
* Accessors		* Accessors
*/		*/
off_t		static off_t
blockif_size(struct blockif_ctxt *bc)		blocklocal_size(struct blockif_ctxt *bc)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_size);		return (bc->bc_size);
}		}

int		static int
blockif_sectsz(struct blockif_ctxt *bc)		blocklocal_sectsz(struct blockif_ctxt *bc)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_sectsz);		return (bc->bc_sectsz);
}		}

void		static void
blockif_psectsz(struct blockif_ctxt bc, int size, int *off)		blocklocal_psectsz(struct blockif_ctxt bc, int size, int *off)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
*size = bc->bc_psectsz;		*size = bc->bc_psectsz;
*off = bc->bc_psectoff;		*off = bc->bc_psectoff;
}		}

int		static int
blockif_queuesz(struct blockif_ctxt *bc)		blocklocal_queuesz(struct blockif_ctxt *bc)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (BLOCKIF_MAXREQ - 1);		return (BLOCKIF_MAXREQ - 1);
}		}

int		static int
blockif_is_ro(struct blockif_ctxt *bc)		blocklocal_is_ro(struct blockif_ctxt *bc)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_rdonly);		return (bc->bc_rdonly);
}		}

int		static int
blockif_candelete(struct blockif_ctxt *bc)		blocklocal_candelete(struct blockif_ctxt *bc)
{		{

assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_candelete);		return (bc->bc_candelete);
}		}

		struct block_backend blocklocal_backend = {
		.bb_name = "file",
		.bb_init = blocklocal_init,
		freqlabsUnsubmitted Done Inline Actions These names seem backwards to me. What do you think about this instead? .bb_name = "blk-local", .bb_scheme = "file", Then name would be the thread name and scheme would be the scheme name for matching the backend. freqlabs: These names seem backwards to me. What do you think about this instead? ``` .bb_name…
		wjw_digiware.nlAuthorUnsubmitted Done Inline Actions This you will have to fight with @trociny, since he did not like the name `scheme`. Note that ATM, `bb_threadinfo` is not used, since we only use what the parent inserted. So it is things like `"%d:%d", pi->pi_slot, pi->pi_func that set the threadname. wjw_digiware.nl: This you will have to fight with @trociny, since he did not like the name `scheme`. Note that…
		trocinyUnsubmitted Done Inline Actions My point actually was why have both bb_name\|prefix and bb_scheme? Why don't just have one param: bb_name, so it is both the thread name and scheme name, whatever. It looks for me perfectly well if one have to specify a device like 'blk-local:/path/to/dev_or_file' trociny: My point actually was why have both bb_name\|prefix and bb_scheme? Why don't just have one param…
		freqlabsUnsubmitted Done Inline Actions I had not noticed that bb_threadinfo isn't used by the blockif abstraction. That being the case, just get rid of it. If a backend needs to store some thread info for its own threads, it can keep it in the private data. With that out of the way I don't mind much whether the remaining member is called bb_name or bb_scheme. freqlabs: I had not noticed that bb_threadinfo isn't used by the blockif abstraction. That being the case…
		.bb_cleanup = blocklocal_cleanup,
		.bb_open = blocklocal_open,
		.bb_size = blocklocal_size,
		.bb_chs = blocklocal_chs,
		.bb_sectsz = blocklocal_sectsz,
		.bb_psectsz = blocklocal_psectsz,
		.bb_queuesz = blocklocal_queuesz,
		.bb_is_ro = blocklocal_is_ro,
		.bb_candelete = blocklocal_candelete,
		.bb_read = blocklocal_read,
		.bb_write = blocklocal_write,
		.bb_flush = blocklocal_flush,
		.bb_delete = blocklocal_delete,
		.bb_cancel = blocklocal_cancel,
		.bb_close = blocklocal_close,
		};
		DATA_SET(block_backend_set, blocklocal_backend);