d/freebsd-dev
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
e8b96c6007
freebsd-dev/module/zfs/zfs_fm.c
Matthew Ahrens e8b96c6007 Illumos #4045 write throttle & i/o scheduler performance work 
4045 zfs write throttle & i/o scheduler performance work

1. The ZFS i/o scheduler (vdev_queue.c) now divides i/os into 5 classes: sync
read, sync write, async read, async write, and scrub/resilver.  The scheduler
issues a number of concurrent i/os from each class to the device.  Once a class
has been selected, an i/o is selected from this class using either an elevator
algorithem (async, scrub classes) or FIFO (sync classes).  The number of
concurrent async write i/os is tuned dynamically based on i/o load, to achieve
good sync i/o latency when there is not a high load of writes, and good write
throughput when there is.  See the block comment in vdev_queue.c (reproduced
below) for more details.

2. The write throttle (dsl_pool_tempreserve_space() and
txg_constrain_throughput()) is rewritten to produce much more consistent delays
when under constant load.  The new write throttle is based on the amount of
dirty data, rather than guesses about future performance of the system.  When
there is a lot of dirty data, each transaction (e.g. write() syscall) will be
delayed by the same small amount.  This eliminates the "brick wall of wait"
that the old write throttle could hit, causing all transactions to wait several
seconds until the next txg opens.  One of the keys to the new write throttle is
decrementing the amount of dirty data as i/o completes, rather than at the end
of spa_sync().  Note that the write throttle is only applied once the i/o
scheduler is issuing the maximum number of outstanding async writes.  See the
block comments in dsl_pool.c and above dmu_tx_delay() (reproduced below) for
more details.

This diff has several other effects, including:

 * the commonly-tuned global variable zfs_vdev_max_pending has been removed;
use per-class zfs_vdev_*_max_active values or zfs_vdev_max_active instead.

 * the size of each txg (meaning the amount of dirty data written, and thus the
time it takes to write out) is now controlled differently.  There is no longer
an explicit time goal; the primary determinant is amount of dirty data.
Systems that are under light or medium load will now often see that a txg is
always syncing, but the impact to performance (e.g. read latency) is minimal.
Tune zfs_dirty_data_max and zfs_dirty_data_sync to control this.

 * zio_taskq_batch_pct = 75 -- Only use 75% of all CPUs for compression,
checksum, etc.  This improves latency by not allowing these CPU-intensive tasks
to consume all CPU (on machines with at least 4 CPU's; the percentage is
rounded up).

--matt

APPENDIX: problems with the current i/o scheduler

The current ZFS i/o scheduler (vdev_queue.c) is deadline based.  The problem
with this is that if there are always i/os pending, then certain classes of
i/os can see very long delays.

For example, if there are always synchronous reads outstanding, then no async
writes will be serviced until they become "past due".  One symptom of this
situation is that each pass of the txg sync takes at least several seconds
(typically 3 seconds).

If many i/os become "past due" (their deadline is in the past), then we must
service all of these overdue i/os before any new i/os.  This happens when we
enqueue a batch of async writes for the txg sync, with deadlines 2.5 seconds in
the future.  If we can't complete all the i/os in 2.5 seconds (e.g. because
there were always reads pending), then these i/os will become past due.  Now we
must service all the "async" writes (which could be hundreds of megabytes)
before we service any reads, introducing considerable latency to synchronous
i/os (reads or ZIL writes).

Notes on porting to ZFS on Linux:

- zio_t gained new members io_physdone and io_phys_children.  Because
  object caches in the Linux port call the constructor only once at
  allocation time, objects may contain residual data when retrieved
  from the cache. Therefore zio_create() was updated to zero out the two
  new fields.

- vdev_mirror_pending() relied on the depth of the per-vdev pending queue
  (vq->vq_pending_tree) to select the least-busy leaf vdev to read from.
  This tree has been replaced by vq->vq_active_tree which is now used
  for the same purpose.

- vdev_queue_init() used the value of zfs_vdev_max_pending to determine
  the number of vdev I/O buffers to pre-allocate.  That global no longer
  exists, so we instead use the sum of the *_max_active values for each of
  the five I/O classes described above.

- The Illumos implementation of dmu_tx_delay() delays a transaction by
  sleeping in condition variable embedded in the thread
  (curthread->t_delay_cv).  We do not have an equivalent CV to use in
  Linux, so this change replaced the delay logic with a wrapper called
  zfs_sleep_until(). This wrapper could be adopted upstream and in other
  downstream ports to abstract away operating system-specific delay logic.

- These tunables are added as module parameters, and descriptions added
  to the zfs-module-parameters.5 man page.

  spa_asize_inflation
  zfs_deadman_synctime_ms
  zfs_vdev_max_active
  zfs_vdev_async_write_active_min_dirty_percent
  zfs_vdev_async_write_active_max_dirty_percent
  zfs_vdev_async_read_max_active
  zfs_vdev_async_read_min_active
  zfs_vdev_async_write_max_active
  zfs_vdev_async_write_min_active
  zfs_vdev_scrub_max_active
  zfs_vdev_scrub_min_active
  zfs_vdev_sync_read_max_active
  zfs_vdev_sync_read_min_active
  zfs_vdev_sync_write_max_active
  zfs_vdev_sync_write_min_active
  zfs_dirty_data_max_percent
  zfs_delay_min_dirty_percent
  zfs_dirty_data_max_max_percent
  zfs_dirty_data_max
  zfs_dirty_data_max_max
  zfs_dirty_data_sync
  zfs_delay_scale

  The latter four have type unsigned long, whereas they are uint64_t in
  Illumos.  This accommodates Linux's module_param() supported types, but
  means they may overflow on 32-bit architectures.

  The values zfs_dirty_data_max and zfs_dirty_data_max_max are the most
  likely to overflow on 32-bit systems, since they express physical RAM
  sizes in bytes.  In fact, Illumos initializes zfs_dirty_data_max_max to
  2^32 which does overflow. To resolve that, this port instead initializes
  it in arc_init() to 25% of physical RAM, and adds the tunable
  zfs_dirty_data_max_max_percent to override that percentage.  While this
  solution doesn't completely avoid the overflow issue, it should be a
  reasonable default for most systems, and the minority of affected
  systems can work around the issue by overriding the defaults.

- Fixed reversed logic in comment above zfs_delay_scale declaration.

- Clarified comments in vdev_queue.c regarding when per-queue minimums take
  effect.

- Replaced dmu_tx_write_limit in the dmu_tx kstat file
  with dmu_tx_dirty_delay and dmu_tx_dirty_over_max.  The first counts
  how many times a transaction has been delayed because the pool dirty
  data has exceeded zfs_delay_min_dirty_percent.  The latter counts how
  many times the pool dirty data has exceeded zfs_dirty_data_max (which
  we expect to never happen).

- The original patch would have regressed the bug fixed in
  zfsonlinux/zfs@c418410, which prevented users from setting the
  zfs_vdev_aggregation_limit tuning larger than SPA_MAXBLOCKSIZE.
  A similar fix is added to vdev_queue_aggregate().

- In vdev_queue_io_to_issue(), dynamically allocate 'zio_t search' on the
  heap instead of the stack.  In Linux we can't afford such large
  structures on the stack.

Reviewed by: George Wilson <george.wilson@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Christopher Siden <christopher.siden@delphix.com>
Reviewed by: Ned Bass <bass6@llnl.gov>
Reviewed by: Brendan Gregg <brendan.gregg@joyent.com>
Approved by: Robert Mustacchi <rm@joyent.com>

References:
  http://www.illumos.org/issues/4045
  illumos/illumos-gate@69962b5647

Ported-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #1913
2013-12-06 09:32:43 -08:00

894 lines
26 KiB
C
Raw Blame History

/*
* CDDL HEADER START
*
* The contents of this file are subject to the terms of the
* Common Development and Distribution License (the "License").
* You may not use this file except in compliance with the License.
*
* You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
* or http://www.opensolaris.org/os/licensing.
* See the License for the specific language governing permissions
* and limitations under the License.
*
* When distributing Covered Code, include this CDDL HEADER in each
* file and include the License file at usr/src/OPENSOLARIS.LICENSE.
* If applicable, add the following below this CDDL HEADER, with the
* fields enclosed by brackets "[]" replaced with your own identifying
* information: Portions Copyright [yyyy] [name of copyright owner]
*
* CDDL HEADER END
*/
/*
* Copyright 2009 Sun Microsystems, Inc. All rights reserved.
* Use is subject to license terms.
*/
/*
* Copyright (c) 2012 by Delphix. All rights reserved.
*/
#include <sys/spa.h>
#include <sys/spa_impl.h>
#include <sys/vdev.h>
#include <sys/vdev_impl.h>
#include <sys/zio.h>
#include <sys/zio_checksum.h>
#include <sys/fm/fs/zfs.h>
#include <sys/fm/protocol.h>
#include <sys/fm/util.h>
#include <sys/sysevent.h>
/*
* This general routine is responsible for generating all the different ZFS
* ereports. The payload is dependent on the class, and which arguments are
* supplied to the function:
*
* EREPORT POOL VDEV IO
* block X X X
* data X X
* device X X
* pool X
*
* If we are in a loading state, all errors are chained together by the same
* SPA-wide ENA (Error Numeric Association).
*
* For isolated I/O requests, we get the ENA from the zio_t. The propagation
* gets very complicated due to RAID-Z, gang blocks, and vdev caching. We want
* to chain together all ereports associated with a logical piece of data. For
* read I/Os, there are basically three 'types' of I/O, which form a roughly
* layered diagram:
*
* +---------------+
* | Aggregate I/O | No associated logical data or device
* +---------------+
* |
* V
* +---------------+ Reads associated with a piece of logical data.
* | Read I/O | This includes reads on behalf of RAID-Z,
* +---------------+ mirrors, gang blocks, retries, etc.
* |
* V
* +---------------+ Reads associated with a particular device, but
* | Physical I/O | no logical data. Issued as part of vdev caching
* +---------------+ and I/O aggregation.
*
* Note that 'physical I/O' here is not the same terminology as used in the rest
* of ZIO. Typically, 'physical I/O' simply means that there is no attached
* blockpointer. But I/O with no associated block pointer can still be related
* to a logical piece of data (i.e. RAID-Z requests).
*
* Purely physical I/O always have unique ENAs. They are not related to a
* particular piece of logical data, and therefore cannot be chained together.
* We still generate an ereport, but the DE doesn't correlate it with any
* logical piece of data. When such an I/O fails, the delegated I/O requests
* will issue a retry, which will trigger the 'real' ereport with the correct
* ENA.
*
* We keep track of the ENA for a ZIO chain through the 'io_logical' member.
* When a new logical I/O is issued, we set this to point to itself. Child I/Os
* then inherit this pointer, so that when it is first set subsequent failures
* will use the same ENA. For vdev cache fill and queue aggregation I/O,
* this pointer is set to NULL, and no ereport will be generated (since it
* doesn't actually correspond to any particular device or piece of data,
* and the caller will always retry without caching or queueing anyway).
*
* For checksum errors, we want to include more information about the actual
* error which occurs. Accordingly, we build an ereport when the error is
* noticed, but instead of sending it in immediately, we hang it off of the
* io_cksum_report field of the logical IO. When the logical IO completes
* (successfully or not), zfs_ereport_finish_checksum() is called with the
* good and bad versions of the buffer (if available), and we annotate the
* ereport with information about the differences.
*/
#ifdef _KERNEL
static void
zfs_zevent_post_cb(nvlist_t *nvl, nvlist_t *detector)
{
if (nvl)
fm_nvlist_destroy(nvl, FM_NVA_FREE);
if (detector)
fm_nvlist_destroy(detector, FM_NVA_FREE);
}
static void
zfs_ereport_start(nvlist_t **ereport_out, nvlist_t **detector_out,
const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
uint64_t stateoroffset, uint64_t size)
{
nvlist_t *ereport, *detector;
uint64_t ena;
char class[64];
/*
* If we are doing a spa_tryimport() or in recovery mode,
* ignore errors.
*/
if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT ||
spa_load_state(spa) == SPA_LOAD_RECOVER)
return;
/*
* If we are in the middle of opening a pool, and the previous attempt
* failed, don't bother logging any new ereports - we're just going to
* get the same diagnosis anyway.
*/
if (spa_load_state(spa) != SPA_LOAD_NONE &&
spa->spa_last_open_failed)
return;
if (zio != NULL) {
/*
* If this is not a read or write zio, ignore the error. This
* can occur if the DKIOCFLUSHWRITECACHE ioctl fails.
*/
if (zio->io_type != ZIO_TYPE_READ &&
zio->io_type != ZIO_TYPE_WRITE)
return;
if (vd != NULL) {
/*
* If the vdev has already been marked as failing due
* to a failed probe, then ignore any subsequent I/O
* errors, as the DE will automatically fault the vdev
* on the first such failure. This also catches cases
* where vdev_remove_wanted is set and the device has
* not yet been asynchronously placed into the REMOVED
* state.
*/
if (zio->io_vd == vd && !vdev_accessible(vd, zio))
return;
/*
* Ignore checksum errors for reads from DTL regions of
* leaf vdevs.
*/
if (zio->io_type == ZIO_TYPE_READ &&
zio->io_error == ECKSUM &&
vd->vdev_ops->vdev_op_leaf &&
vdev_dtl_contains(vd, DTL_MISSING, zio->io_txg, 1))
return;
}
}
/*
* For probe failure, we want to avoid posting ereports if we've
* already removed the device in the meantime.
*/
if (vd != NULL &&
strcmp(subclass, FM_EREPORT_ZFS_PROBE_FAILURE) == 0 &&
(vd->vdev_remove_wanted || vd->vdev_state == VDEV_STATE_REMOVED))
return;
if ((ereport = fm_nvlist_create(NULL)) == NULL)
return;
if ((detector = fm_nvlist_create(NULL)) == NULL) {
fm_nvlist_destroy(ereport, FM_NVA_FREE);
return;
}
/*
* Serialize ereport generation
*/
mutex_enter(&spa->spa_errlist_lock);
/*
* Determine the ENA to use for this event. If we are in a loading
* state, use a SPA-wide ENA. Otherwise, if we are in an I/O state, use
* a root zio-wide ENA. Otherwise, simply use a unique ENA.
*/
if (spa_load_state(spa) != SPA_LOAD_NONE) {
if (spa->spa_ena == 0)
spa->spa_ena = fm_ena_generate(0, FM_ENA_FMT1);
ena = spa->spa_ena;
} else if (zio != NULL && zio->io_logical != NULL) {
if (zio->io_logical->io_ena == 0)
zio->io_logical->io_ena =
fm_ena_generate(0, FM_ENA_FMT1);
ena = zio->io_logical->io_ena;
} else {
ena = fm_ena_generate(0, FM_ENA_FMT1);
}
/*
* Construct the full class, detector, and other standard FMA fields.
*/
(void) snprintf(class, sizeof (class), "%s.%s",
ZFS_ERROR_CLASS, subclass);
fm_fmri_zfs_set(detector, FM_ZFS_SCHEME_VERSION, spa_guid(spa),
vd != NULL ? vd->vdev_guid : 0);
fm_ereport_set(ereport, FM_EREPORT_VERSION, class, ena, detector, NULL);
/*
* Construct the per-ereport payload, depending on which parameters are
* passed in.
*/
/*
* Generic payload members common to all ereports.
*/
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL,
DATA_TYPE_STRING, spa_name(spa), FM_EREPORT_PAYLOAD_ZFS_POOL_GUID,
DATA_TYPE_UINT64, spa_guid(spa),
FM_EREPORT_PAYLOAD_ZFS_POOL_CONTEXT, DATA_TYPE_INT32,
spa_load_state(spa), NULL);
if (spa != NULL) {
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_POOL_FAILMODE,
DATA_TYPE_STRING,
spa_get_failmode(spa) == ZIO_FAILURE_MODE_WAIT ?
FM_EREPORT_FAILMODE_WAIT :
spa_get_failmode(spa) == ZIO_FAILURE_MODE_CONTINUE ?
FM_EREPORT_FAILMODE_CONTINUE : FM_EREPORT_FAILMODE_PANIC,
NULL);
}
if (vd != NULL) {
vdev_t *pvd = vd->vdev_parent;
vdev_queue_t *vq = &vd->vdev_queue;
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID,
DATA_TYPE_UINT64, vd->vdev_guid,
FM_EREPORT_PAYLOAD_ZFS_VDEV_TYPE,
DATA_TYPE_STRING, vd->vdev_ops->vdev_op_type, NULL);
if (vd->vdev_path != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_PATH,
DATA_TYPE_STRING, vd->vdev_path, NULL);
if (vd->vdev_devid != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_DEVID,
DATA_TYPE_STRING, vd->vdev_devid, NULL);
if (vd->vdev_fru != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_FRU,
DATA_TYPE_STRING, vd->vdev_fru, NULL);
if (vd->vdev_ashift)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_ASHIFT,
DATA_TYPE_UINT64, vd->vdev_ashift, NULL);
if (vq != NULL) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_COMP_TS,
DATA_TYPE_UINT64, vq->vq_io_complete_ts, NULL);
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_VDEV_DELTA_TS,
DATA_TYPE_UINT64, vq->vq_io_delta_ts, NULL);
}
if (pvd != NULL) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PARENT_GUID,
DATA_TYPE_UINT64, pvd->vdev_guid,
FM_EREPORT_PAYLOAD_ZFS_PARENT_TYPE,
DATA_TYPE_STRING, pvd->vdev_ops->vdev_op_type,
NULL);
if (pvd->vdev_path)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PARENT_PATH,
DATA_TYPE_STRING, pvd->vdev_path, NULL);
if (pvd->vdev_devid)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PARENT_DEVID,
DATA_TYPE_STRING, pvd->vdev_devid, NULL);
}
}
if (zio != NULL) {
/*
* Payload common to all I/Os.
*/
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_ERR,
DATA_TYPE_INT32, zio->io_error, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_FLAGS,
DATA_TYPE_INT32, zio->io_flags, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_STAGE,
DATA_TYPE_UINT32, zio->io_stage, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_PIPELINE,
DATA_TYPE_UINT32, zio->io_pipeline, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELAY,
DATA_TYPE_UINT64, zio->io_delay, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_TIMESTAMP,
DATA_TYPE_UINT64, zio->io_timestamp, NULL);
fm_payload_set(ereport, FM_EREPORT_PAYLOAD_ZFS_ZIO_DELTA,
DATA_TYPE_UINT64, zio->io_delta, NULL);
/*
* If the 'size' parameter is non-zero, it indicates this is a
* RAID-Z or other I/O where the physical offset and length are
* provided for us, instead of within the zio_t.
*/
if (vd != NULL) {
if (size)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
DATA_TYPE_UINT64, stateoroffset,
FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
DATA_TYPE_UINT64, size, NULL);
else
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OFFSET,
DATA_TYPE_UINT64, zio->io_offset,
FM_EREPORT_PAYLOAD_ZFS_ZIO_SIZE,
DATA_TYPE_UINT64, zio->io_size, NULL);
}
/*
* Payload for I/Os with corresponding logical information.
*/
if (zio->io_logical != NULL)
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJSET,
DATA_TYPE_UINT64,
zio->io_logical->io_bookmark.zb_objset,
FM_EREPORT_PAYLOAD_ZFS_ZIO_OBJECT,
DATA_TYPE_UINT64,
zio->io_logical->io_bookmark.zb_object,
FM_EREPORT_PAYLOAD_ZFS_ZIO_LEVEL,
DATA_TYPE_INT64,
zio->io_logical->io_bookmark.zb_level,
FM_EREPORT_PAYLOAD_ZFS_ZIO_BLKID,
DATA_TYPE_UINT64,
zio->io_logical->io_bookmark.zb_blkid, NULL);
} else if (vd != NULL) {
/*
* If we have a vdev but no zio, this is a device fault, and the
* 'stateoroffset' parameter indicates the previous state of the
* vdev.
*/
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_PREV_STATE,
DATA_TYPE_UINT64, stateoroffset, NULL);
}
mutex_exit(&spa->spa_errlist_lock);
*ereport_out = ereport;
*detector_out = detector;
}
/* if it's <= 128 bytes, save the corruption directly */
#define ZFM_MAX_INLINE (128 / sizeof (uint64_t))
#define MAX_RANGES 16
typedef struct zfs_ecksum_info {
/* histograms of set and cleared bits by bit number in a 64-bit word */
uint16_t zei_histogram_set[sizeof (uint64_t) * NBBY];
uint16_t zei_histogram_cleared[sizeof (uint64_t) * NBBY];
/* inline arrays of bits set and cleared. */
uint64_t zei_bits_set[ZFM_MAX_INLINE];
uint64_t zei_bits_cleared[ZFM_MAX_INLINE];
/*
* for each range, the number of bits set and cleared. The Hamming
* distance between the good and bad buffers is the sum of them all.
*/
uint32_t zei_range_sets[MAX_RANGES];
uint32_t zei_range_clears[MAX_RANGES];
struct zei_ranges {
uint32_t zr_start;
uint32_t zr_end;
} zei_ranges[MAX_RANGES];
size_t zei_range_count;
uint32_t zei_mingap;
uint32_t zei_allowed_mingap;
} zfs_ecksum_info_t;
static void
update_histogram(uint64_t value_arg, uint16_t *hist, uint32_t *count)
{
size_t i;
size_t bits = 0;
uint64_t value = BE_64(value_arg);
/* We store the bits in big-endian (largest-first) order */
for (i = 0; i < 64; i++) {
if (value & (1ull << i)) {
hist[63 - i]++;
++bits;
}
}
/* update the count of bits changed */
*count += bits;
}
/*
* We've now filled up the range array, and need to increase "mingap" and
* shrink the range list accordingly. zei_mingap is always the smallest
* distance between array entries, so we set the new_allowed_gap to be
* one greater than that. We then go through the list, joining together
* any ranges which are closer than the new_allowed_gap.
*
* By construction, there will be at least one. We also update zei_mingap
* to the new smallest gap, to prepare for our next invocation.
*/
static void
zei_shrink_ranges(zfs_ecksum_info_t *eip)
{
uint32_t mingap = UINT32_MAX;
uint32_t new_allowed_gap = eip->zei_mingap + 1;
size_t idx, output;
size_t max = eip->zei_range_count;
struct zei_ranges *r = eip->zei_ranges;
ASSERT3U(eip->zei_range_count, >, 0);
ASSERT3U(eip->zei_range_count, <=, MAX_RANGES);
output = idx = 0;
while (idx < max - 1) {
uint32_t start = r[idx].zr_start;
uint32_t end = r[idx].zr_end;
while (idx < max - 1) {
uint32_t nstart, nend, gap;
idx++;
nstart = r[idx].zr_start;
nend = r[idx].zr_end;
gap = nstart - end;
if (gap < new_allowed_gap) {
end = nend;
continue;
}
if (gap < mingap)
mingap = gap;
break;
}
r[output].zr_start = start;
r[output].zr_end = end;
output++;
}
ASSERT3U(output, <, eip->zei_range_count);
eip->zei_range_count = output;
eip->zei_mingap = mingap;
eip->zei_allowed_mingap = new_allowed_gap;
}
static void
zei_add_range(zfs_ecksum_info_t *eip, int start, int end)
{
struct zei_ranges *r = eip->zei_ranges;
size_t count = eip->zei_range_count;
if (count >= MAX_RANGES) {
zei_shrink_ranges(eip);
count = eip->zei_range_count;
}
if (count == 0) {
eip->zei_mingap = UINT32_MAX;
eip->zei_allowed_mingap = 1;
} else {
int gap = start - r[count - 1].zr_end;
if (gap < eip->zei_allowed_mingap) {
r[count - 1].zr_end = end;
return;
}
if (gap < eip->zei_mingap)
eip->zei_mingap = gap;
}
r[count].zr_start = start;
r[count].zr_end = end;
eip->zei_range_count++;
}
static size_t
zei_range_total_size(zfs_ecksum_info_t *eip)
{
struct zei_ranges *r = eip->zei_ranges;
size_t count = eip->zei_range_count;
size_t result = 0;
size_t idx;
for (idx = 0; idx < count; idx++)
result += (r[idx].zr_end - r[idx].zr_start);
return (result);
}
static zfs_ecksum_info_t *
annotate_ecksum(nvlist_t *ereport, zio_bad_cksum_t *info,
const uint8_t *goodbuf, const uint8_t *badbuf, size_t size,
boolean_t drop_if_identical)
{
const uint64_t *good = (const uint64_t *)goodbuf;
const uint64_t *bad = (const uint64_t *)badbuf;
uint64_t allset = 0;
uint64_t allcleared = 0;
size_t nui64s = size / sizeof (uint64_t);
size_t inline_size;
int no_inline = 0;
size_t idx;
size_t range;
size_t offset = 0;
ssize_t start = -1;
zfs_ecksum_info_t *eip = kmem_zalloc(sizeof (*eip), KM_PUSHPAGE);
/* don't do any annotation for injected checksum errors */
if (info != NULL && info->zbc_injected)
return (eip);
if (info != NULL && info->zbc_has_cksum) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_EXPECTED,
DATA_TYPE_UINT64_ARRAY,
sizeof (info->zbc_expected) / sizeof (uint64_t),
(uint64_t *)&info->zbc_expected,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_ACTUAL,
DATA_TYPE_UINT64_ARRAY,
sizeof (info->zbc_actual) / sizeof (uint64_t),
(uint64_t *)&info->zbc_actual,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_ALGO,
DATA_TYPE_STRING,
info->zbc_checksum_name,
NULL);
if (info->zbc_byteswapped) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_CKSUM_BYTESWAP,
DATA_TYPE_BOOLEAN, 1,
NULL);
}
}
if (badbuf == NULL || goodbuf == NULL)
return (eip);
ASSERT3U(nui64s, <=, UINT16_MAX);
ASSERT3U(size, ==, nui64s * sizeof (uint64_t));
ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
ASSERT3U(size, <=, UINT32_MAX);
/* build up the range list by comparing the two buffers. */
for (idx = 0; idx < nui64s; idx++) {
if (good[idx] == bad[idx]) {
if (start == -1)
continue;
zei_add_range(eip, start, idx);
start = -1;
} else {
if (start != -1)
continue;
start = idx;
}
}
if (start != -1)
zei_add_range(eip, start, idx);
/* See if it will fit in our inline buffers */
inline_size = zei_range_total_size(eip);
if (inline_size > ZFM_MAX_INLINE)
no_inline = 1;
/*
* If there is no change and we want to drop if the buffers are
* identical, do so.
*/
if (inline_size == 0 && drop_if_identical) {
kmem_free(eip, sizeof (*eip));
return (NULL);
}
/*
* Now walk through the ranges, filling in the details of the
* differences. Also convert our uint64_t-array offsets to byte
* offsets.
*/
for (range = 0; range < eip->zei_range_count; range++) {
size_t start = eip->zei_ranges[range].zr_start;
size_t end = eip->zei_ranges[range].zr_end;
for (idx = start; idx < end; idx++) {
uint64_t set, cleared;
// bits set in bad, but not in good
set = ((~good[idx]) & bad[idx]);
// bits set in good, but not in bad
cleared = (good[idx] & (~bad[idx]));
allset |= set;
allcleared |= cleared;
if (!no_inline) {
ASSERT3U(offset, <, inline_size);
eip->zei_bits_set[offset] = set;
eip->zei_bits_cleared[offset] = cleared;
offset++;
}
update_histogram(set, eip->zei_histogram_set,
&eip->zei_range_sets[range]);
update_histogram(cleared, eip->zei_histogram_cleared,
&eip->zei_range_clears[range]);
}
/* convert to byte offsets */
eip->zei_ranges[range].zr_start *= sizeof (uint64_t);
eip->zei_ranges[range].zr_end *= sizeof (uint64_t);
}
eip->zei_allowed_mingap *= sizeof (uint64_t);
inline_size *= sizeof (uint64_t);
/* fill in ereport */
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_BAD_OFFSET_RANGES,
DATA_TYPE_UINT32_ARRAY, 2 * eip->zei_range_count,
(uint32_t *)eip->zei_ranges,
FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_MIN_GAP,
DATA_TYPE_UINT32, eip->zei_allowed_mingap,
FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_SETS,
DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_sets,
FM_EREPORT_PAYLOAD_ZFS_BAD_RANGE_CLEARS,
DATA_TYPE_UINT32_ARRAY, eip->zei_range_count, eip->zei_range_clears,
NULL);
if (!no_inline) {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_BAD_SET_BITS,
DATA_TYPE_UINT8_ARRAY,
inline_size, (uint8_t *)eip->zei_bits_set,
FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_BITS,
DATA_TYPE_UINT8_ARRAY,
inline_size, (uint8_t *)eip->zei_bits_cleared,
NULL);
} else {
fm_payload_set(ereport,
FM_EREPORT_PAYLOAD_ZFS_BAD_SET_HISTOGRAM,
DATA_TYPE_UINT16_ARRAY,
NBBY * sizeof (uint64_t), eip->zei_histogram_set,
FM_EREPORT_PAYLOAD_ZFS_BAD_CLEARED_HISTOGRAM,
DATA_TYPE_UINT16_ARRAY,
NBBY * sizeof (uint64_t), eip->zei_histogram_cleared,
NULL);
}
return (eip);
}
#endif
void
zfs_ereport_post(const char *subclass, spa_t *spa, vdev_t *vd, zio_t *zio,
uint64_t stateoroffset, uint64_t size)
{
#ifdef _KERNEL
nvlist_t *ereport = NULL;
nvlist_t *detector = NULL;
zfs_ereport_start(&ereport, &detector,
subclass, spa, vd, zio, stateoroffset, size);
if (ereport == NULL)
return;
/* Cleanup is handled by the callback function */
zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
#endif
}
void
zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd,
struct zio *zio, uint64_t offset, uint64_t length, void *arg,
zio_bad_cksum_t *info)
{
zio_cksum_report_t *report = kmem_zalloc(sizeof (*report), KM_PUSHPAGE);
if (zio->io_vsd != NULL)
zio->io_vsd_ops->vsd_cksum_report(zio, report, arg);
else
zio_vsd_default_cksum_report(zio, report, arg);
/* copy the checksum failure information if it was provided */
if (info != NULL) {
report->zcr_ckinfo = kmem_zalloc(sizeof (*info), KM_PUSHPAGE);
bcopy(info, report->zcr_ckinfo, sizeof (*info));
}
report->zcr_align = 1ULL << vd->vdev_top->vdev_ashift;
report->zcr_length = length;
#ifdef _KERNEL
zfs_ereport_start(&report->zcr_ereport, &report->zcr_detector,
FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio, offset, length);
if (report->zcr_ereport == NULL) {
report->zcr_free(report->zcr_cbdata, report->zcr_cbinfo);
if (report->zcr_ckinfo != NULL) {
kmem_free(report->zcr_ckinfo,
sizeof (*report->zcr_ckinfo));
}
kmem_free(report, sizeof (*report));
return;
}
#endif
mutex_enter(&spa->spa_errlist_lock);
report->zcr_next = zio->io_logical->io_cksum_report;
zio->io_logical->io_cksum_report = report;
mutex_exit(&spa->spa_errlist_lock);
}
void
zfs_ereport_finish_checksum(zio_cksum_report_t *report,
const void *good_data, const void *bad_data, boolean_t drop_if_identical)
{
#ifdef _KERNEL
zfs_ecksum_info_t *info = NULL;
info = annotate_ecksum(report->zcr_ereport, report->zcr_ckinfo,
good_data, bad_data, report->zcr_length, drop_if_identical);
if (info != NULL)
zfs_zevent_post(report->zcr_ereport,
report->zcr_detector, zfs_zevent_post_cb);
report->zcr_ereport = report->zcr_detector = NULL;
if (info != NULL)
kmem_free(info, sizeof (*info));
#endif
}
void
zfs_ereport_free_checksum(zio_cksum_report_t *rpt)
{
#ifdef _KERNEL
if (rpt->zcr_ereport != NULL) {
fm_nvlist_destroy(rpt->zcr_ereport,
FM_NVA_FREE);
fm_nvlist_destroy(rpt->zcr_detector,
FM_NVA_FREE);
}
#endif
rpt->zcr_free(rpt->zcr_cbdata, rpt->zcr_cbinfo);
if (rpt->zcr_ckinfo != NULL)
kmem_free(rpt->zcr_ckinfo, sizeof (*rpt->zcr_ckinfo));
kmem_free(rpt, sizeof (*rpt));
}
void
zfs_ereport_send_interim_checksum(zio_cksum_report_t *report)
{
#ifdef _KERNEL
zfs_zevent_post(report->zcr_ereport, report->zcr_detector, NULL);
#endif
}
void
zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd,
struct zio *zio, uint64_t offset, uint64_t length,
const void *good_data, const void *bad_data, zio_bad_cksum_t *zbc)
{
#ifdef _KERNEL
nvlist_t *ereport = NULL;
nvlist_t *detector = NULL;
zfs_ecksum_info_t *info;
zfs_ereport_start(&ereport, &detector,
FM_EREPORT_ZFS_CHECKSUM, spa, vd, zio, offset, length);
if (ereport == NULL)
return;
info = annotate_ecksum(ereport, zbc, good_data, bad_data, length,
B_FALSE);
if (info != NULL) {
zfs_zevent_post(ereport, detector, zfs_zevent_post_cb);
kmem_free(info, sizeof (*info));
}
#endif
}
static void
zfs_post_common(spa_t *spa, vdev_t *vd, const char *name)
{
#ifdef _KERNEL
nvlist_t *resource;
char class[64];
if (spa_load_state(spa) == SPA_LOAD_TRYIMPORT)
return;
if ((resource = fm_nvlist_create(NULL)) == NULL)
return;
(void) snprintf(class, sizeof (class), "%s.%s.%s", FM_RSRC_RESOURCE,
ZFS_ERROR_CLASS, name);
VERIFY(nvlist_add_uint8(resource, FM_VERSION, FM_RSRC_VERSION) == 0);
VERIFY(nvlist_add_string(resource, FM_CLASS, class) == 0);
VERIFY(nvlist_add_uint64(resource,
FM_EREPORT_PAYLOAD_ZFS_POOL_GUID, spa_guid(spa)) == 0);
if (vd) {
VERIFY(nvlist_add_uint64(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_GUID, vd->vdev_guid) == 0);
VERIFY(nvlist_add_uint64(resource,
FM_EREPORT_PAYLOAD_ZFS_VDEV_STATE, vd->vdev_state) == 0);
}
zfs_zevent_post(resource, NULL, zfs_zevent_post_cb);
#endif
}
/*
* The 'resource.fs.zfs.removed' event is an internal signal that the given vdev
* has been removed from the system. This will cause the DE to ignore any
* recent I/O errors, inferring that they are due to the asynchronous device
* removal.
*/
void
zfs_post_remove(spa_t *spa, vdev_t *vd)
{
zfs_post_common(spa, vd, FM_EREPORT_RESOURCE_REMOVED);
}
/*
* The 'resource.fs.zfs.autoreplace' event is an internal signal that the pool
* has the 'autoreplace' property set, and therefore any broken vdevs will be
* handled by higher level logic, and no vdev fault should be generated.
*/
void
zfs_post_autoreplace(spa_t *spa, vdev_t *vd)
{
zfs_post_common(spa, vd, FM_EREPORT_RESOURCE_AUTOREPLACE);
}
/*
* The 'resource.fs.zfs.statechange' event is an internal signal that the
* given vdev has transitioned its state to DEGRADED or HEALTHY. This will
* cause the retire agent to repair any outstanding fault management cases
* open because the device was not found (fault.fs.zfs.device).
*/
void
zfs_post_state_change(spa_t *spa, vdev_t *vd)
{
zfs_post_common(spa, vd, FM_EREPORT_RESOURCE_STATECHANGE);
}
#if defined(_KERNEL) && defined(HAVE_SPL)
EXPORT_SYMBOL(zfs_ereport_post);
EXPORT_SYMBOL(zfs_ereport_post_checksum);
EXPORT_SYMBOL(zfs_post_remove);
EXPORT_SYMBOL(zfs_post_autoreplace);
EXPORT_SYMBOL(zfs_post_state_change);
#endif /* _KERNEL */
Reference in New Issue View Git Blame Copy Permalink