Differential D23010 Diff 67204 usr.sbin/bhyve/block_if.c

Changeset View

Standalone View

usr.sbin/bhyve/block_if.c

This file was copied to usr.sbin/bhyve/block_backends.c.

Show First 20 Lines • Show All 52 Lines • ▼ Show 20 Lines
#include <pthread.h>		#include <pthread.h>
#include <pthread_np.h>		#include <pthread_np.h>
#include <signal.h>		#include <signal.h>
#include <sysexits.h>		#include <sysexits.h>
#include <unistd.h>		#include <unistd.h>

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

		#include <sys/linker_set.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		SET_DECLARE(block_backend_set, block_backend_t);

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

enum blockop {		struct locblk_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 locblk_sig_elem *bse_next;
};		};

static struct blockif_sig_elem *blockif_bse_head;		static struct locblk_sig_elem *locblk_bse_head;

static int		static int
blockif_enqueue(struct blockif_ctxt bc, struct blockif_req breq,		locblk_enqueue(struct locblk_ctxt bc, struct blockif_req breq,
enum blockop op)		enum blockop op)
{		{
struct blockif_elem be, tbe;		struct locblk_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);
assert(be->be_status == BST_FREE);		assert(be->be_status == BST_FREE);
TAILQ_REMOVE(&bc->bc_freeq, be, be_link);		TAILQ_REMOVE(&bc->bc_freeq, be, be_link);
be->be_req = breq;		be->be_req = breq;
Show All 24 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)		locblk_dequeue(struct locblk_ctxt bc, pthread_t t, struct locblk_elem *bep)
{		{
struct blockif_elem *be;		struct locblk_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)		locblk_complete(struct locblk_ctxt bc, struct locblk_elem be)
{		{
struct blockif_elem *tbe;		struct locblk_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)		locblk_proc(struct locblk_ctxt bc, struct locblk_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)
▲ Show 20 Lines • Show All 104 Lines • ▼ Show 20 Lines	locblk_proc(struct locblk_ctxt bc, struct locblk_elem be, uint8_t *buf)
}		}

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)		locblk_thr(void *arg)
{		{
struct blockif_ctxt *bc;		struct locblk_ctxt *bc;
struct blockif_elem *be;		struct locblk_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 (locblk_dequeue(bc, t, &be)) {
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);
blockif_proc(bc, be, buf);		locblk_proc(bc, be, buf);
pthread_mutex_lock(&bc->bc_mtx);		pthread_mutex_lock(&bc->bc_mtx);
blockif_complete(bc, be);		locblk_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)		locblk_sigcont_handler(int signal, enum ev_type type, void *arg)
{		{
struct blockif_sig_elem *bse;		struct locblk_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 = locblk_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 *)&locblk_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 int
		locblk_request(struct locblk_ctxt bc, struct blockif_req breq,
		enum blockop op)
		{
		int err = 0;

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

		return (err);
		}

		/*
		* Routines for access blockdevices on the local system.
		* be it either a file, or a block device
		*/
static void		static void
blockif_init(void)		locblk_init(void)
{		{
mevent_add(SIGCONT, EVF_SIGNAL, blockif_sigcont_handler, NULL);		mevent_add(SIGCONT, EVF_SIGNAL, locblk_sigcont_handler, NULL);
(void) signal(SIGCONT, SIG_IGN);		(void) signal(SIGCONT, SIG_IGN);
}		}

struct blockif_ctxt *		static void
blockif_open(const char optstr, const char ident)		locblk_cleanup(block_backend_t **be)
		{ /* empty block
		* currently no cleanup required.
		*/
		}

		static block_backend_t locblk_backend;

		static int
		locblk_open(block_backend_t *be, const char optstr, const char *pci_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 locblk_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
		freqlabsUnsubmitted Done Inline Actions Indentation looks different than in the rest of the file. freqlabs: Indentation looks different than in the rest of the file.

pthread_once(&blockif_once, blockif_init);		pthread_once(&locblk_once, locblk_init);

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

/*		/*
Show All 22 Lines	#endif
}		}

extra = 0;		extra = 0;
if (nocache)		if (nocache)
extra \|= O_DIRECT;		extra \|= O_DIRECT;
if (sync)		if (sync)
extra \|= O_SYNC;		extra \|= O_SYNC;

		/*
		* device and option parsing completed correctly so set the
		* backend return value to this backend.
		* Anything that returns an error below is due to errors in
		* handling of the device. But this is the correct backend.
		*/
		*be = &locblk_backend;

fd = open(nopt, (ro ? O_RDONLY : O_RDWR) \| extra);		fd = open(nopt, (ro ? O_RDONLY : O_RDWR) \| extra);
if (fd < 0 && !ro) {		if (fd < 0 && !ro) {
/* Attempt a r/w fail with a r/o open */		/* Attempt a r/w fail with a r/o open */
fd = open(nopt, O_RDONLY \| extra);		fd = open(nopt, O_RDONLY \| extra);
ro = 1;		ro = 1;
}		}

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

		freqlabsUnsubmitted Done Inline Actions If you null out the ':' before moving the pointer forward then you can use strcmp below to ensure we get an exact match for the backend name. optrest = '\0'; optrest++; freqlabs:* If you null out the ':' before moving the pointer forward then you can use strcmp below to…
#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 All 25 Lines	if (ssopt != 0) {
}		}

/*		/*
* Some backend drivers (e.g. cd0, ada0) require that the I/O		* Some backend drivers (e.g. cd0, ada0) require that the I/O
* size be a multiple of the device's sector size.		* size be a multiple of the device's sector size.
*		*
* Validate that the emulated sector size complies with this		* Validate that the emulated sector size complies with this
* requirement.		* requirement.
*/		*/
if (S_ISCHR(sbuf.st_mode)) {		if (S_ISCHR(sbuf.st_mode)) {
if (ssopt < sectsz \|\| (ssopt % sectsz) != 0) {		if (ssopt < sectsz \|\| (ssopt % sectsz) != 0) {
EPRINTLN("Sector size %d incompatible "		EPRINTLN("Sector size %d incompatible "
"with underlying device sector size %d",		"with underlying device sector size %d",
ssopt, sectsz);		ssopt, sectsz);
		freqlabsUnsubmitted Done Inline Actions This comment is outdated. freqlabs: This comment is outdated.
goto err;		goto err;
}		}
}		}

sectsz = ssopt;		sectsz = ssopt;
psectsz = pssopt;		psectsz = pssopt;
		freqlabsUnsubmitted Done Inline Actions We can `break` if open fails. There should be only one backend with the given name. freqlabs: We can `break` if open fails. There should be only one backend with the given name.
psectoff = 0;		psectoff = 0;
}		}

bc = calloc(1, sizeof(struct blockif_ctxt));		bc = calloc(1, sizeof(struct locblk_ctxt));
if (bc == NULL) {		if (bc == NULL) {
perror("calloc");		perror("calloc");
goto err;		goto err;
}		}
		/* And update the backend descriptor for this */
		(be)->bc = (struct blockif_ctxt)bc;

bc->bc_magic = BLOCKIF_SIG;		bc->bc_magic = BLOCKIF_SIG;
bc->bc_fd = fd;		bc->bc_fd = fd;
bc->bc_ischr = S_ISCHR(sbuf.st_mode);		bc->bc_ischr = S_ISCHR(sbuf.st_mode);
		trocinyUnsubmitted Done Inline Actions Note, bb_open may return NULL here and you will be dereferencing NULL trociny: Note, bb_open may return NULL here and you will be dereferencing NULL
		wjw_digiware.nlAuthorUnsubmitted Done Inline Actions good call wjw_digiware.nl: good call
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, locblk_thr, bc);
snprintf(tname, sizeof(tname), "blk-%s-%d", ident, i);		snprintf(tname, sizeof(tname), "blk-%s-%d", pci_ident, i);
pthread_set_name_np(bc->bc_btid[i], tname);		pthread_set_name_np(bc->bc_btid[i], tname);
}		}

return (bc);		return (0);
err:		err:
if (fd >= 0)		if (fd >= 0)
close(fd);		close(fd);
free(nopt);		free(nopt);
return (NULL);		return (-1);
}		}

		trocinyUnsubmitted Done Inline Actions What is a point of passing the scheme prefix to the driver? The first step that it would need to do is to trim it. Why not instead always pass to the bb_open optstr without scheme. It is easy to trim it from the optstr that has it -- just increase optstr by the prefix length. trociny: What is a point of passing the scheme prefix to the driver? The first step that it would need…
		wjw_digiware.nlAuthorUnsubmitted Done Inline Actions Yeah, that is the other way around ..... Makes it easier here, but a bit more code in the `FOR loop` wjw_digiware.nl: Yeah, that is the other way around ..... Makes it easier here, but a bit more code in the `FOR…
		trocinyUnsubmitted Done Inline Actions Otherwise you would have "a bit mode code" in every driver to parse the prefix. trociny: Otherwise you would have "a bit mode code" in every driver to parse the prefix.
static int
blockif_request(struct blockif_ctxt bc, struct blockif_req breq,
enum blockop op)
{
int err;

err = 0;

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

return (err);
}

int		int
		trocinyUnsubmitted Done Inline Actions BTW, in my version I just check `optstr` for a prefix ending with semicolon (e.g. 'rbd:'), and then use this prefix to find the driver. trociny: BTW, in my version I just check `optstr` for a prefix ending with semicolon (e.g. 'rbd:'), and…
		wjw_digiware.nlAuthorUnsubmitted Done Inline Actions @trociny I think it is beter actually to leave this up to the backends themselfs, this creates more flexibility. And would for example allow for multiple `/........` backends. Where if the basic local backend refuses, there could be other backends working on this `optstr` that can handle the request. wjw_digiware.nl: @trociny I think it is beter actually to leave this up to the backends themselfs, this…
		trocinyUnsubmitted Done Inline Actions I can hardly imagine how backends are going to detect if its "their" device, if only they would not rely on some prefix name. And if they would why not to do this in generic way instead of implementing this in every driver? So, bhyve would look for provided disk name, and if it had "foo:" prefix, it would assume that the driver is "foo", and would try to open the device using this driver (if it had been successfully loaded. trociny: I can hardly imagine how backends are going to detect if its "their" device, if only they would…
		freqlabsUnsubmitted Done Inline Actions I agree with `scheme:` suggest, ie `file://`, `rbd://`, `nbd://`, etc. with a fallback to the file backend if no scheme is given. Each backend can register its scheme name and we can directly pick the right one with no ambiguity. freqlabs: I agree with `scheme:` suggest, ie `file://`, `rbd://`, `nbd://`, etc. with a fallback to the…
		freqlabsUnsubmitted Done Inline Actions Oops, I meant to remove `//` from the rest of those before I submit. Anything after the `:` would be passed to the backend. freqlabs: Oops, I meant to remove `//` from the rest of those before I submit. Anything after the `:`…
		wjw_digiware.nlAuthorUnsubmitted Done Inline Actions Right, and perhaps keep the `/......` for legacy defaulting to local store. wjw_digiware.nl: Right, and perhaps keep the `/......` for legacy defaulting to local store.
		wjw_digiware.nlAuthorUnsubmitted Done Inline Actions @trociny I guess it would be possible to add a scheme descriptor to the backend struct, and only go with that. But then still you'd have to call `bb_open()` to see it the `optstr` is a valid string and attach the device. wjw_digiware.nl: @trociny I guess it would be possible to add a scheme descriptor to the backend struct, and…
blockif_read(struct blockif_ctxt bc, struct blockif_req breq)		locblk_read(block_backend_t be, struct blockif_req breq)
{		{
		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (blockif_request(bc, breq, BOP_READ));		return (locblk_request(bc, breq, BOP_READ));
}		}

int		int
blockif_write(struct blockif_ctxt bc, struct blockif_req breq)		locblk_write(block_backend_t be, struct blockif_req breq)
{		{

		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
		freqlabsUnsubmitted Done Inline Actions `!= NULL` please to avoid confusion about the type. freqlabs: `!= NULL` please to avoid confusion about the type.
return (blockif_request(bc, breq, BOP_WRITE));		return (locblk_request(bc, breq, BOP_WRITE));
}		}

int		int
blockif_flush(struct blockif_ctxt bc, struct blockif_req breq)		locblk_flush(block_backend_t be, struct blockif_req breq)
{		{

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

int		int
blockif_delete(struct blockif_ctxt bc, struct blockif_req breq)		locblk_delete(block_backend_t be, struct blockif_req breq)
{		{

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

int		int
blockif_cancel(struct blockif_ctxt bc, struct blockif_req breq)		locblk_cancel(block_backend_t be, struct blockif_req breq)
{		{
struct blockif_elem *be;		struct locblk_elem *belem;

		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
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(belem, &bc->bc_pendq, be_link) {
if (be->be_req == breq)		if (belem->be_req == breq)
break;		break;
}		}
if (be != NULL) {		if (be != NULL) {
/*		/*
* Found it.		* Found it.
*/		*/
blockif_complete(bc, be);		locblk_complete(bc, belem);
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);

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

/*		/*
* Check in-flight requests.		* Check in-flight requests.
*/		*/
TAILQ_FOREACH(be, &bc->bc_busyq, be_link) {		TAILQ_FOREACH(belem, &bc->bc_busyq, be_link) {
if (be->be_req == breq)		if (belem->be_req == breq)
break;		break;
}		}
if (be == NULL) {		if (belem == NULL) {
/*		/*
* Didn't find it.		* Didn't find it.
*/		*/
pthread_mutex_unlock(&bc->bc_mtx);		pthread_mutex_unlock(&bc->bc_mtx);
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 (belem->be_status == BST_BUSY) {
struct blockif_sig_elem bse, *old_head;		struct locblk_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 = locblk_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 *)&locblk_bse_head,
(uintptr_t)old_head,		(uintptr_t)old_head,
(uintptr_t)&bse));		(uintptr_t)&bse));

pthread_kill(be->be_tid, SIGCONT);		pthread_kill(belem->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		int
blockif_close(struct blockif_ctxt *bc)		locblk_close(block_backend_t *be)
{		{
void *jval;		void *jval;
int i;		int i;

		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
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(bc->bc_fd);
free(bc);		free(bc);

return (0);		return (0);
		freqlabsUnsubmitted Done Inline Actions This isn't the only place I see this, but I'll just point it out here: Minor style nit - `Values in return statements should be enclosed in parentheses.` return (bc->be->close(bc)); freqlabs: This isn't the only place I see this, but I'll just point it out here: Minor style nit…
		wjw_digiware.nlAuthorUnsubmitted Done Inline Actions I have not fixed this in files I have not touched. wjw_digiware.nl: I have not fixed this in files I have not touched.
}		}

/*		/*
* 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		void
blockif_chs(struct blockif_ctxt bc, uint16_t c, uint8_t h, uint8_t s)		locblk_chs(block_backend_t be, 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;

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

sectors = bc->bc_size / bc->bc_sectsz;		sectors = bc->bc_size / bc->bc_sectsz;

/* Clamp the size to the largest possible with CHS */		/* Clamp the size to the largest possible with CHS */
if (sectors > 65535UL16255)		if (sectors > 65535UL16255)
sectors = 65535UL16255;		sectors = 65535UL16255;

Show All 25 Lines	locblk_chs(block_backend_t be, uint16_t c, uint8_t h, uint8_t s)
*h = heads;		*h = heads;
*s = secpt;		*s = secpt;
}		}

/*		/*
* Accessors		* Accessors
*/		*/
off_t		off_t
blockif_size(struct blockif_ctxt *bc)		locblk_size(block_backend_t *be)
{		{
		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_size);		return (bc->bc_size);
}		}

int		int
blockif_sectsz(struct blockif_ctxt *bc)		locblk_sectsz(block_backend_t *be)
{		{
		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_sectsz);		return (bc->bc_sectsz);
}		}

void		void
blockif_psectsz(struct blockif_ctxt bc, int size, int *off)		locblk_psectsz(block_backend_t be, int size, int *off)
{		{
		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
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		int
blockif_queuesz(struct blockif_ctxt *bc)		locblk_queuesz(block_backend_t *be)
{		{
		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (BLOCKIF_MAXREQ - 1);		return (BLOCKIF_MAXREQ - 1);
}		}

int		int
blockif_is_ro(struct blockif_ctxt *bc)		locblk_is_ro(block_backend_t *be)
{		{
		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_rdonly);		return (bc->bc_rdonly);
}		}

int		int
blockif_candelete(struct blockif_ctxt *bc)		locblk_candelete(block_backend_t *be)
{		{
		locblk_ctxt_t bc = (locblk_ctxt_t )(be->bc);
assert(bc->bc_magic == BLOCKIF_SIG);		assert(bc->bc_magic == BLOCKIF_SIG);
return (bc->bc_candelete);		return (bc->bc_candelete);
}		}

		static
		block_backend_t locblk_backend = {
		.prefix = "blk-local",
		.init = locblk_init,
		.cleanup = locblk_cleanup,
		.open = locblk_open,
		.size = locblk_size,
		.chs = locblk_chs,
		.sectsz = locblk_sectsz,
		.psectsz = locblk_psectsz,
		.queuesz = locblk_queuesz,
		.is_ro = locblk_is_ro,
		.candelete = locblk_candelete,
		.read = locblk_read,
		.write = locblk_write,
		.flush = locblk_flush,
		.delete = locblk_delete,
		.cancel = locblk_cancel,
		.close = locblk_close,
		.priv_size = 0,
		};
		DATA_SET(block_backend_set, locblk_backend);