Archive-Team
/
zfs
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 103 Wiki Activity

Merge commit 'refs/top-bases/linux-zfs-branch' into linux-zfs-branch

This commit is contained in:
Brian Behlendorf 2009-07-29 17:53:18 -07:00
parent acbe64c078 95198f4d9b
commit 04da8a9177
9 changed files with 313 additions and 180 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
config/kernel-bio-args.m4
View File

@ -1,24 +0,0 @@
dnl #
dnl # 2.6.x API change
dnl # bio_end_io_t uses 2 args (size was dropped from prototype)
dnl #
AC_DEFUN([ZFS_AC_CONFIG_KERNEL_BIO_ARGS],
[AC_MSG_CHECKING([whether bio_end_io_t wants 2 args])
tmp_flags="$EXTRA_KCFLAGS"
EXTRA_KCFLAGS="-Werror"
ZFS_LINUX_TRY_COMPILE([
#include <linux/bio.h>
],[
void (*wanted_end_io)(struct bio *, int) = NULL;
bio_end_io_t *local_end_io;
local_end_io = wanted_end_io;
],[
AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_2ARGS_BIO_END_IO_T, 1,
[bio_end_io_t wants 2 args])
],[
AC_MSG_RESULT(no)
])
EXTRA_KCFLAGS="$tmp_flags"
])

20
config/kernel-bio-empty-barrier.m4 Normal file
View File

@ -0,0 +1,20 @@
dnl #
dnl # 2.6.24 API change
dnl # Empty write barriers are now supported and we should use them.
dnl #
AC_DEFUN([ZFS_AC_KERNEL_BIO_EMPTY_BARRIER], [
AC_MSG_CHECKING([whether bio_empty_barrier() is defined])
EXTRA_KCFLAGS="-Werror"
ZFS_LINUX_TRY_COMPILE([
#include <linux/bio.h>
],[
struct bio bio;
(void)bio_empty_barrier(&bio);
],[
AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_BIO_EMPTY_BARRIER, 1,
[bio_empy_barrier() is defined])
],[
AC_MSG_RESULT(no)
])
])

29
config/kernel-bio-end-io-t-args.m4 Normal file
View File

@ -0,0 +1,29 @@
dnl #
dnl # 2.6.24 API change
dnl # Size argument dropped from bio_endio and bi_end_io, because the
dnl # bi_end_io is only called once now when the request is complete.
dnl # There is no longer any need for a size argument. This also means
dnl # that partial IO's are no longer possibe and the end_io callback
dnl # should not check bi->bi_size. Finally, the return type was updated
dnl # to void.
dnl #
AC_DEFUN([ZFS_AC_KERNEL_BIO_END_IO_T_ARGS], [
AC_MSG_CHECKING([whether bio_end_io_t wants 2 args])
tmp_flags="$EXTRA_KCFLAGS"
EXTRA_KCFLAGS="-Werror"
ZFS_LINUX_TRY_COMPILE([
#include <linux/bio.h>
],[
void (*wanted_end_io)(struct bio *, int) = NULL;
bio_end_io_t *local_end_io;
local_end_io = wanted_end_io;
],[
AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_2ARGS_BIO_END_IO_T, 1,
[bio_end_io_t wants 2 args])
],[
AC_MSG_RESULT(no)
])
EXTRA_KCFLAGS="$tmp_flags"
])

18
config/kernel-bio-rw-syncio.m4 Normal file
View File

@ -0,0 +1,18 @@
dnl #
dnl # 2.6.29 API change
dnl # BIO_RW_SYNC renamed to BIO_RW_SYNCIO
dnl #
AC_DEFUN([ZFS_AC_KERNEL_BIO_RW_SYNCIO], [
AC_MSG_CHECKING([whether BIO_RW_SYNCIO is defined])
ZFS_LINUX_TRY_COMPILE([
#include <linux/bio.h>
],[
int flags = BIO_RW_SYNCIO;
],[
AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_BIO_RW_SYNCIO, 1,
[BIO_RW_SYNCIO is defined])
],[
AC_MSG_RESULT(no)
])
])

19
config/kernel-invalidate-bdev-args.m4 Normal file
View File

@ -0,0 +1,19 @@
dnl #
dnl # 2.6.22 API change
dnl # Unused destroy_dirty_buffers arg removed from prototype.
dnl #
AC_DEFUN([ZFS_AC_KERNEL_INVALIDATE_BDEV_ARGS], [
AC_MSG_CHECKING([whether invalidate_bdev() wants 1 arg])
ZFS_LINUX_TRY_COMPILE([
#include <linux/buffer_head.h>
],[
struct block_device *bdev;
invalidate_bdev(bdev);
],[
AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_1ARG_INVALIDATE_BDEV, 1,
[invalidate_bdev() wants 1 arg])
],[
AC_MSG_RESULT(no)
])
])

12
config/kernel-open-bdev-exclusive.m4 Normal file
View File

@ -0,0 +1,12 @@
dnl #
dnl # 2.6.28 API change
dnl # open/close_bdev_excl() renamed to open/close_bdev_exclusive()
dnl #
AC_DEFUN([ZFS_AC_KERNEL_OPEN_BDEV_EXCLUSIVE], [
ZFS_CHECK_SYMBOL_EXPORT(
[open_bdev_exclusive],
[fs/block_dev.c],
[AC_DEFINE(HAVE_OPEN_BDEV_EXCLUSIVE, 1,
[open_bdev_exclusive() is available])],
[])
])

6
config/kernel.m4
View File

@ -4,7 +4,11 @@ dnl #
AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [ AC_DEFUN([ZFS_AC_CONFIG_KERNEL], [
ZFS_AC_KERNEL ZFS_AC_KERNEL
ZFS_AC_SPL ZFS_AC_SPL
ZFS_AC_CONFIG_KERNEL_BIO_ARGS ZFS_AC_KERNEL_OPEN_BDEV_EXCLUSIVE
ZFS_AC_KERNEL_INVALIDATE_BDEV_ARGS
ZFS_AC_KERNEL_BIO_END_IO_T_ARGS
ZFS_AC_KERNEL_BIO_RW_SYNCIO
ZFS_AC_KERNEL_BIO_EMPTY_BARRIER
dnl # Kernel build make options dnl # Kernel build make options
dnl # KERNELMAKE_PARAMS="V=1" # Enable verbose module build dnl # KERNELMAKE_PARAMS="V=1" # Enable verbose module build

40
module/zfs/include/sys/vdev_disk.h
View File

@ -10,15 +10,51 @@ extern "C" {
#include <sys/ddi.h> #include <sys/ddi.h>
#include <sys/sunldi.h> #include <sys/sunldi.h>
#include <sys/sunddi.h> #include <sys/sunddi.h>
#include <zfs_config.h>
typedef struct vdev_disk { typedef struct vdev_disk {
ddi_devid_t vd_devid; ddi_devid_t vd_devid;
char *vd_minor; char *vd_minor;
ldi_handle_t vd_lh; struct block_device *vd_bdev;
} vdev_disk_t; } vdev_disk_t;
extern int vdev_disk_physio(ldi_handle_t, caddr_t, size_t, uint64_t, int); extern int vdev_disk_physio(struct block_device *, caddr_t,
size_t, uint64_t, int);
extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **); extern int vdev_disk_read_rootlabel(char *, char *, nvlist_t **);
/* 2.6.24 API change */
#ifdef HAVE_2ARGS_BIO_END_IO_T
# define BIO_END_IO_PROTO(fn, x, y, z) static void fn(struct bio *x, int z)
# define BIO_END_IO_RETURN(rc) return
#else
# define BIO_END_IO_PROTO(fn, x, y, z) static int fn(struct bio *x, \
unsigned int y, int z)
# define BIO_END_IO_RETURN(rc) return rc
#endif /* HAVE_2ARGS_BIO_END_IO_T */
/* 2.6.29 API change */
#ifdef HAVE_BIO_RW_SYNCIO
# define DIO_RW_SYNCIO BIO_RW_SYNCIO
#else
# define DIO_RW_SYNCIO BIO_RW_SYNC
#endif /* HAVE_BIO_RW_SYNCIO */
/* 2.6.28 API change */
#ifdef HAVE_OPEN_BDEV_EXCLUSIVE
# define vdev_bdev_open(path, md, hld) open_bdev_exclusive(path, md, hld)
# define vdev_bdev_close(bdev, md) close_bdev_exclusive(bdev, md)
#else
# define vdev_bdev_open(path, md, hld) open_bdev_excl(path, md, hld)
# define vdev_bdev_close(bdev, md) close_bdev_excl(bdev)
#endif /* HAVE_OPEN_BDEV_EXCLUSIVE */
/* 2.6.22 API change */
#ifdef HAVE_1ARG_INVALIDATE_BDEV
# define vdev_bdev_invalidate(bdev) invalidate_bdev(bdev)
#else
# define vdev_bdev_invalidate(bdev) invalidate_bdev(bdev, 1)
#endif /* HAVE_1ARG_INVALIDATE_BDEV */
#endif /* _KERNEL */ #endif /* _KERNEL */
#ifdef __cplusplus #ifdef __cplusplus

319
module/zfs/vdev_disk.c
View File

@ -19,7 +19,7 @@
* CDDL HEADER END * CDDL HEADER END
*/ */
/* /*
* Copyright 2007 Sun Microsystems, Inc. All rights reserved. * Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms. * Use is subject to license terms.
*/ */
@ -30,7 +30,6 @@
#include <sys/fs/zfs.h> #include <sys/fs/zfs.h>
#include <sys/zio.h> #include <sys/zio.h>
#include <sys/sunldi.h> #include <sys/sunldi.h>
#include <zfs_config.h>
/* /*
* Virtual device vector for disks. * Virtual device vector for disks.
@ -46,58 +45,90 @@ typedef struct dio_request {
struct bio *dr_bio[0]; /* Attached bio's */ struct bio *dr_bio[0]; /* Attached bio's */
} dio_request_t; } dio_request_t;
static int
vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift) #ifdef HAVE_OPEN_BDEV_EXCLUSIVE
static fmode_t
vdev_bdev_mode(int smode)
{ {
struct block_device *vd_lh; fmode_t mode = 0;
vdev_disk_t *dvd;
ASSERT3S(smode & (FREAD | FWRITE), !=, 0);
if (smode & FREAD)
mode |= FMODE_READ;
if (smode & FWRITE)
mode |= FMODE_WRITE;
return mode;
}
#else
static int
vdev_bdev_mode(int smode)
{
int mode = 0;
ASSERT3S(smode & (FREAD | FWRITE), !=, 0);
if ((smode & FREAD) && !(smode & FWRITE))
mode = MS_RDONLY;
return mode;
}
#endif /* HAVE_OPEN_BDEV_EXCLUSIVE */
static int
vdev_disk_open(vdev_t *v, uint64_t *psize, uint64_t *ashift)
{
struct block_device *bdev;
vdev_disk_t *vd;
int mode;
/* Must have a pathname and it must be absolute. */ /* Must have a pathname and it must be absolute. */
if (vd->vdev_path == NULL || vd->vdev_path[0] != '/') { if (v->vdev_path == NULL || v->vdev_path[0] != '/') {
vd->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL; v->vdev_stat.vs_aux = VDEV_AUX_BAD_LABEL;
return EINVAL; return EINVAL;
} }
dvd = kmem_zalloc(sizeof(vdev_disk_t), KM_SLEEP); vd = kmem_zalloc(sizeof(vdev_disk_t), KM_SLEEP);
if (dvd == NULL) if (vd == NULL)
return ENOMEM; return ENOMEM;
/* XXX: Since we do not have devid support like Solaris we /*
* XXX: Since we do not have devid support like Solaris we
* currently can't be as clever about opening the right device. * currently can't be as clever about opening the right device.
* For now we will simply open the device name provided and * For now we will simply open the device name provided and
* fail when it doesn't exist. If your devices get reordered * fail when it doesn't exist. If your devices get reordered
* your going to be screwed, use udev for now to prevent this. * your going to be screwed, use udev for now to prevent this.
*
* XXX: mode here could be the global spa_mode with a little
* munging of the flags to make then more agreeable to linux.
* However, simply passing a 0 for now gets us W/R behavior.
*/ */
vd_lh = open_bdev_excl(vd->vdev_path, 0, dvd); mode = spa_mode(v->vdev_spa);
if (IS_ERR(vd_lh)) { bdev = vdev_bdev_open(v->vdev_path, vdev_bdev_mode(mode), vd);
kmem_free(dvd, sizeof(vdev_disk_t)); if (IS_ERR(bdev)) {
return -PTR_ERR(vd_lh); kmem_free(vd, sizeof(vdev_disk_t));
return -PTR_ERR(bdev);
} }
/* XXX: Long term validate stored dvd->vd_devid with a unique /*
* XXX: Long term validate stored vd->vd_devid with a unique
* identifier read from the disk, likely EFI support. * identifier read from the disk, likely EFI support.
*/ */
vd->vdev_tsd = dvd; v->vdev_tsd = vd;
dvd->vd_lh = vd_lh; vd->vd_bdev = bdev;
/* Check if this is a whole device. When vd_lh->bd_contains == /* Check if this is a whole device. When bdev->bd_contains ==
* vd_lh we have a whole device and not simply a partition. */ * bdev we have a whole device and not simply a partition. */
vd->vdev_wholedisk = !!(vd_lh->bd_contains == vd_lh); v->vdev_wholedisk = !!(bdev->bd_contains == bdev);
/* Clear the nowritecache bit, causes vdev_reopen() to try again. */ /* Clear the nowritecache bit, causes vdev_reopen() to try again. */
vd->vdev_nowritecache = B_FALSE; v->vdev_nowritecache = B_FALSE;
/* Determine the actual size of the device (in bytes) /* Determine the actual size of the device (in bytes)
* *
* XXX: SECTOR_SIZE is defined to 512b which may not be true for * XXX: SECTOR_SIZE is defined to 512b which may not be true for
* your device, we must use the actual hardware sector size. * your device, we must use the actual hardware sector size.
*/ */
*psize = get_capacity(vd_lh->bd_disk) * SECTOR_SIZE; *psize = get_capacity(bdev->bd_disk) * SECTOR_SIZE;
/* Based on the minimum sector size set the block size */ /* Based on the minimum sector size set the block size */
*ashift = highbit(MAX(SECTOR_SIZE, SPA_MINBLOCKSIZE)) - 1; *ashift = highbit(MAX(SECTOR_SIZE, SPA_MINBLOCKSIZE)) - 1;
@ -106,27 +137,22 @@ vdev_disk_open(vdev_t *vd, uint64_t *psize, uint64_t *ashift)
} }
static void static void
vdev_disk_close(vdev_t *vd) vdev_disk_close(vdev_t *v)
{ {
vdev_disk_t *dvd = vd->vdev_tsd; vdev_disk_t *vd = v->vdev_tsd;
if (dvd == NULL) if (vd == NULL)
return; return;
if (dvd->vd_lh != NULL) if (vd->vd_bdev != NULL)
close_bdev_excl(dvd->vd_lh); vdev_bdev_close(vd->vd_bdev,
vdev_bdev_mode(spa_mode(v->vdev_spa)));
kmem_free(dvd, sizeof(vdev_disk_t)); kmem_free(vd, sizeof(vdev_disk_t));
vd->vdev_tsd = NULL; v->vdev_tsd = NULL;
} }
#ifdef HAVE_2ARGS_BIO_END_IO_T BIO_END_IO_PROTO(vdev_disk_physio_completion, bio, size, rc)
static void
vdev_disk_physio_completion(struct bio *bio, int rc)
#else
static int
vdev_disk_physio_completion(struct bio *bio, unsigned int size, int rc)
#endif /* HAVE_2ARGS_BIO_END_IO_T */
{ {
dio_request_t *dr = bio->bi_private; dio_request_t *dr = bio->bi_private;
zio_t *zio; zio_t *zio;
@ -168,7 +194,7 @@ vdev_disk_physio_completion(struct bio *bio, unsigned int size, int rc)
spin_unlock(&dr->dr_lock); spin_unlock(&dr->dr_lock);
/* Synchronous dio cleanup handled by waiter */ /* Synchronous dio cleanup handled by waiter */
if (dr->dr_rw & (1 << BIO_RW_SYNC)) { if (dr->dr_rw & (1 << DIO_RW_SYNCIO)) {
complete(&dr->dr_comp); complete(&dr->dr_comp);
} else { } else {
for (i = 0; i < dr->dr_bio_count; i++) for (i = 0; i < dr->dr_bio_count; i++)
@ -186,13 +212,7 @@ vdev_disk_physio_completion(struct bio *bio, unsigned int size, int rc)
spin_unlock(&dr->dr_lock); spin_unlock(&dr->dr_lock);
} }
rc = 0; BIO_END_IO_RETURN(0);
#ifdef HAVE_2ARGS_BIO_END_IO_T
return;
#else
return rc;
#endif /* HAVE_2ARGS_BIO_END_IO_T */
} }
static struct bio * static struct bio *
@ -250,16 +270,19 @@ bio_map(struct request_queue *q, void *data, unsigned int len, gfp_t gfp_mask)
} }
static int static int
__vdev_disk_physio(struct block_device *vd_lh, zio_t *zio, caddr_t kbuf_ptr, __vdev_disk_physio(struct block_device *bdev, zio_t *zio, caddr_t kbuf_ptr,
size_t kbuf_size, uint64_t kbuf_offset, int flags) size_t kbuf_size, uint64_t kbuf_offset, int flags)
{ {
struct request_queue *q = vd_lh->bd_disk->queue; struct request_queue *q;
dio_request_t *dr; dio_request_t *dr;
caddr_t bio_ptr; caddr_t bio_ptr;
uint64_t bio_offset; uint64_t bio_offset;
int i, j, error = 0, bio_count, bio_size, dio_size; int i, j, error = 0, bio_count, bio_size, dio_size;
ASSERT3S(kbuf_offset % SECTOR_SIZE, ==, 0); ASSERT3S(kbuf_offset % SECTOR_SIZE, ==, 0);
q = bdev_get_queue(bdev);
if (!q)
return ENXIO;
bio_count = (kbuf_size / (q->max_hw_sectors << 9)) + 1; bio_count = (kbuf_size / (q->max_hw_sectors << 9)) + 1;
dio_size = sizeof(dio_request_t) + sizeof(struct bio *) * bio_count; dio_size = sizeof(dio_request_t) + sizeof(struct bio *) * bio_count;
@ -271,13 +294,10 @@ __vdev_disk_physio(struct block_device *vd_lh, zio_t *zio, caddr_t kbuf_ptr,
spin_lock_init(&dr->dr_lock); spin_lock_init(&dr->dr_lock);
dr->dr_ref = 0; dr->dr_ref = 0;
dr->dr_zio = zio; dr->dr_zio = zio;
dr->dr_rw = READ; dr->dr_rw = flags;
dr->dr_error = 0; dr->dr_error = 0;
dr->dr_bio_count = bio_count; dr->dr_bio_count = bio_count;
if (flags & (1 << BIO_RW))
dr->dr_rw = (flags & (1 << BIO_RW_SYNC)) ? WRITE_SYNC : WRITE;
#ifdef BIO_RW_FAILFAST #ifdef BIO_RW_FAILFAST
if (flags & (1 << BIO_RW_FAILFAST)) if (flags & (1 << BIO_RW_FAILFAST))
dr->dr_rw |= 1 << BIO_RW_FAILFAST; dr->dr_rw |= 1 << BIO_RW_FAILFAST;
@ -305,7 +325,7 @@ __vdev_disk_physio(struct block_device *vd_lh, zio_t *zio, caddr_t kbuf_ptr,
return error; return error;
} }
dr->dr_bio[i]->bi_bdev = vd_lh; dr->dr_bio[i]->bi_bdev = bdev;
dr->dr_bio[i]->bi_sector = bio_offset >> 9; dr->dr_bio[i]->bi_sector = bio_offset >> 9;
dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion; dr->dr_bio[i]->bi_end_io = vdev_disk_physio_completion;
dr->dr_bio[i]->bi_private = dr; dr->dr_bio[i]->bi_private = dr;
@ -320,12 +340,14 @@ __vdev_disk_physio(struct block_device *vd_lh, zio_t *zio, caddr_t kbuf_ptr,
submit_bio(dr->dr_rw, dr->dr_bio[i]); submit_bio(dr->dr_rw, dr->dr_bio[i]);
/* /*
* On syncronous blocking requests we wait for all bio the completion * On synchronous blocking requests we wait for all bio the completion
* callbacks to run. We will be woken when the last callback runs * callbacks to run. We will be woken when the last callback runs
* for this dio. We are responsible for freeing the dio_request_t as * for this dio. We are responsible for freeing the dio_request_t as
* well as the final reference on all attached bios. * well as the final reference on all attached bios. Currently, the
* only synchronous consumer is vdev_disk_read_rootlabel() all other
* IO originating from vdev_disk_io_start() is asynchronous.
*/ */
if (dr->dr_rw & (1 << BIO_RW_SYNC)) { if (dr->dr_rw & (1 << DIO_RW_SYNCIO)) {
wait_for_completion(&dr->dr_comp); wait_for_completion(&dr->dr_comp);
ASSERT(dr->dr_ref == 0); ASSERT(dr->dr_ref == 0);
error = dr->dr_error; error = dr->dr_error;
@ -340,87 +362,90 @@ __vdev_disk_physio(struct block_device *vd_lh, zio_t *zio, caddr_t kbuf_ptr,
} }
int int
vdev_disk_physio(ldi_handle_t vd_lh, caddr_t kbuf, vdev_disk_physio(struct block_device *bdev, caddr_t kbuf,
size_t size, uint64_t offset, int flags) size_t size, uint64_t offset, int flags)
{ {
return __vdev_disk_physio(vd_lh, NULL, kbuf, size, offset, flags); return __vdev_disk_physio(bdev, NULL, kbuf, size, offset, flags);
} }
#if 0 /* 2.6.24 API change */
/* XXX: Not yet supported */ #ifdef HAVE_BIO_EMPTY_BARRIER
static void BIO_END_IO_PROTO(vdev_disk_io_flush_completion, bio, size, rc)
vdev_disk_ioctl_done(void *zio_arg, int error)
{ {
zio_t *zio = zio_arg; zio_t *zio = bio->bi_private;
zio->io_error = error; zio->io_error = -rc;
if (rc && (rc == -EOPNOTSUPP))
zio->io_vd->vdev_nowritecache = B_TRUE;
bio_put(bio);
zio_interrupt(zio); zio_interrupt(zio);
BIO_END_IO_RETURN(0);
} }
#endif
static int
vdev_disk_io_flush(struct block_device *bdev, zio_t *zio)
{
struct request_queue *q;
struct bio *bio;
q = bdev_get_queue(bdev);
if (!q)
return ENXIO;
bio = bio_alloc(GFP_KERNEL, 0);
if (!bio)
return ENOMEM;
bio->bi_end_io = vdev_disk_io_flush_completion;
bio->bi_private = zio;
bio->bi_bdev = bdev;
submit_bio(WRITE_BARRIER, bio);
return 0;
}
#else
static int
vdev_disk_io_flush(struct block_device *bdev, zio_t *zio)
{
return ENOTSUP;
}
#endif /* HAVE_BIO_EMPTY_BARRIER */
static int static int
vdev_disk_io_start(zio_t *zio) vdev_disk_io_start(zio_t *zio)
{ {
vdev_t *vd = zio->io_vd; vdev_t *v = zio->io_vd;
vdev_disk_t *dvd = vd->vdev_tsd; vdev_disk_t *vd = v->vdev_tsd;
int flags, error; int flags, error;
if (zio->io_type == ZIO_TYPE_IOCTL) { switch (zio->io_type) {
case ZIO_TYPE_IOCTL:
/* XXPOLICY */ if (!vdev_readable(v)) {
if (!vdev_readable(vd)) {
zio->io_error = ENXIO; zio->io_error = ENXIO;
return ZIO_PIPELINE_CONTINUE; return ZIO_PIPELINE_CONTINUE;
} }
switch (zio->io_cmd) { switch (zio->io_cmd) {
case DKIOCFLUSHWRITECACHE: case DKIOCFLUSHWRITECACHE:
if (zfs_nocacheflush) if (zfs_nocacheflush)
break; break;
if (vd->vdev_nowritecache) { if (v->vdev_nowritecache) {
zio->io_error = ENOTSUP; zio->io_error = ENOTSUP;
break; break;
} }
#if 0 error = vdev_disk_io_flush(vd->vd_bdev, zio);
/* XXX: Not yet supported */ if (error == 0)
vdev_disk_t *dvd = vd->vdev_tsd;
zio->io_dk_callback.dkc_callback = vdev_disk_ioctl_done;
zio->io_dk_callback.dkc_flag = FLUSH_VOLATILE;
zio->io_dk_callback.dkc_cookie = zio;
error = ldi_ioctl(dvd->vd_lh, zio->io_cmd,
(uintptr_t)&zio->io_dk_callback,
FKIOCTL, kcred, NULL);
if (error == 0) {
/*
* The ioctl will be done asychronously,
* and will call vdev_disk_ioctl_done()
* upon completion.
*/
return ZIO_PIPELINE_STOP; return ZIO_PIPELINE_STOP;
}
#else
error = ENOTSUP;
#endif
if (error == ENOTSUP || error == ENOTTY) {
/*
* If we get ENOTSUP or ENOTTY, we know that
* no future attempts will ever succeed.
* In this case we set a persistent bit so
* that we don't bother with the ioctl in the
* future.
*/
vd->vdev_nowritecache = B_TRUE;
}
zio->io_error = error; zio->io_error = error;
if (error == ENOTSUP)
v->vdev_nowritecache = B_TRUE;
break; break;
@ -429,20 +454,30 @@ vdev_disk_io_start(zio_t *zio)
} }
return ZIO_PIPELINE_CONTINUE; return ZIO_PIPELINE_CONTINUE;
case ZIO_TYPE_WRITE:
flags = WRITE;
break;
case ZIO_TYPE_READ:
if (zio->io_flags & ZIO_FLAG_SPECULATIVE)
flags = READA;
else
flags = READ;
break;
default:
zio->io_error = ENOTSUP;
return ZIO_PIPELINE_CONTINUE;
} }
/*
* B_BUSY XXX: Not supported
* B_NOCACHE XXX: Not supported
*/
flags = ((zio->io_type == ZIO_TYPE_READ) ? READ : WRITE);
#ifdef BIO_RW_FAILFAST #ifdef BIO_RW_FAILFAST
if (zio->io_flags & ZIO_FLAG_IO_RETRY) if (zio->io_flags & (ZIO_FLAG_IO_RETRY | ZIO_FLAG_TRYHARD))
flags |= (1 << BIO_RW_FAILFAST); flags |= (1 << BIO_RW_FAILFAST);
#endif /* BIO_RW_FAILFAST */ #endif /* BIO_RW_FAILFAST */
error = __vdev_disk_physio(dvd->vd_lh, zio, zio->io_data, error = __vdev_disk_physio(vd->vd_bdev, zio, zio->io_data,
zio->io_size, zio->io_offset, flags); zio->io_size, zio->io_offset, flags);
if (error) { if (error) {
zio->io_error = error; zio->io_error = error;
@ -456,22 +491,20 @@ static void
vdev_disk_io_done(zio_t *zio) vdev_disk_io_done(zio_t *zio)
{ {
/* /*
* If the device returned EIO, then attempt a DKIOCSTATE ioctl to see if * If the device returned EIO, we revalidate the media. If it is
* the device has been removed. If this is the case, then we trigger an * determined the media has changed this triggers the asynchronous
* asynchronous removal of the device. Otherwise, probe the device and * removal of the device from the configuration.
* make sure it's still accessible.
*/ */
VERIFY3S(zio->io_error, ==, 0); if (zio->io_error == EIO) {
#if 0 vdev_t *v = zio->io_vd;
vdev_disk_t *dvd = vd->vdev_tsd; vdev_disk_t *vd = v->vdev_tsd;
int state = DKIO_NONE;
if (ldi_ioctl(dvd->vd_lh, DKIOCSTATE, (intptr_t)&state, if (check_disk_change(vd->vd_bdev)) {
FKIOCTL, kcred, NULL) == 0 && state != DKIO_INSERTED) { vdev_bdev_invalidate(vd->vd_bdev);
vd->vdev_remove_wanted = B_TRUE; v->vdev_remove_wanted = B_TRUE;
spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE); spa_async_request(zio->io_spa, SPA_ASYNC_REMOVE);
} }
#endif }
} }
vdev_ops_t vdev_disk_ops = { vdev_ops_t vdev_disk_ops = {
@ -492,32 +525,18 @@ vdev_ops_t vdev_disk_ops = {
int int
vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config) vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config)
{ {
struct block_device *vd_lh; struct block_device *bdev;
vdev_label_t *label; vdev_label_t *label;
uint64_t s, size; uint64_t s, size;
int i; int i;
/* bdev = vdev_bdev_open(devpath, vdev_bdev_mode(FREAD), NULL);
* Read the device label and build the nvlist. if (IS_ERR(bdev))
* XXX: Not yet supported return -PTR_ERR(bdev);
*/
#if 0
if (devid != NULL && ddi_devid_str_decode(devid, &tmpdevid,
&minor_name) == 0) {
error = ldi_open_by_devid(tmpdevid, minor_name, spa_mode,
kcred, &vd_lh, zfs_li);
ddi_devid_free(tmpdevid);
ddi_devid_str_free(minor_name);
}
#endif
vd_lh = open_bdev_excl(devpath, MS_RDONLY, NULL); s = get_capacity(bdev->bd_disk) * SECTOR_SIZE;
if (IS_ERR(vd_lh))
return -PTR_ERR(vd_lh);
s = get_capacity(vd_lh->bd_disk) * SECTOR_SIZE;
if (s == 0) { if (s == 0) {
close_bdev_excl(vd_lh); vdev_bdev_close(bdev, vdev_bdev_mode(FREAD));
return EIO; return EIO;
} }
@ -529,8 +548,8 @@ vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config)
/* read vdev label */ /* read vdev label */
offset = vdev_label_offset(size, i, 0); offset = vdev_label_offset(size, i, 0);
if (vdev_disk_physio(vd_lh, (caddr_t)label, if (vdev_disk_physio(bdev, (caddr_t)label,
VDEV_SKIP_SIZE + VDEV_PHYS_SIZE, offset, READ) != 0) VDEV_SKIP_SIZE + VDEV_PHYS_SIZE, offset, READ_SYNC) != 0)
continue; continue;
if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist, if (nvlist_unpack(label->vl_vdev_phys.vp_nvlist,
@ -557,7 +576,7 @@ vdev_disk_read_rootlabel(char *devpath, char *devid, nvlist_t **config)
} }
vmem_free(label, sizeof(vdev_label_t)); vmem_free(label, sizeof(vdev_label_t));
close_bdev_excl(vd_lh); vdev_bdev_close(bdev, vdev_bdev_mode(FREAD));
return 0; return 0;
} }