Reimplement nvd(4) detach handling.

Previous code typically crashed in case of NVMe device unplug or even clean
detach while some I/Os are still in flight.  To fix this the new code calls
disk_gone() and waits for confirmation of all references gone before calling
disk_destroy(), freeing other resources and allowing controller detach.

While there, fix disk lists locking and reimplement unit numbers assignment.

MFC after:	1 month
Sponsored by:	iXsystems, Inc.
This commit is contained in:
Alexander Motin 2018-12-27 18:28:19 +00:00
parent 9fbbfa5d7f
commit 4d07b68988

View file

@ -2,6 +2,7 @@
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (C) 2012-2016 Intel Corporation
* Copyright (C) 2018 Alexander Motin <mav@FreeBSD.org>
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
@ -34,9 +35,11 @@ __FBSDID("$FreeBSD$");
#include <sys/kernel.h>
#include <sys/malloc.h>
#include <sys/module.h>
#include <sys/queue.h>
#include <sys/sysctl.h>
#include <sys/systm.h>
#include <sys/taskqueue.h>
#include <machine/atomic.h>
#include <geom/geom.h>
#include <geom/geom_disk.h>
@ -46,15 +49,16 @@ __FBSDID("$FreeBSD$");
#define NVD_STR "nvd"
struct nvd_disk;
struct nvd_controller;
static disk_ioctl_t nvd_ioctl;
static disk_strategy_t nvd_strategy;
static dumper_t nvd_dump;
static void nvd_done(void *arg, const struct nvme_completion *cpl);
static void nvd_gone(struct nvd_disk *ndisk);
static void *nvd_new_disk(struct nvme_namespace *ns, void *ctrlr);
static void destroy_geom_disk(struct nvd_disk *ndisk);
static void *nvd_new_controller(struct nvme_controller *ctrlr);
static void nvd_controller_fail(void *ctrlr);
@ -67,6 +71,7 @@ MALLOC_DEFINE(M_NVD, "nvd", "nvd(4) allocations");
struct nvme_consumer *consumer_handle;
struct nvd_disk {
struct nvd_controller *ctrlr;
struct bio_queue_head bioq;
struct task bioqtask;
@ -78,6 +83,7 @@ struct nvd_disk {
uint32_t cur_depth;
uint32_t ordered_in_flight;
u_int unit;
TAILQ_ENTRY(nvd_disk) global_tailq;
TAILQ_ENTRY(nvd_disk) ctrlr_tailq;
@ -89,6 +95,7 @@ struct nvd_controller {
TAILQ_HEAD(, nvd_disk) disk_head;
};
static struct mtx nvd_lock;
static TAILQ_HEAD(, nvd_controller) ctrlr_head;
static TAILQ_HEAD(disk_list, nvd_disk) disk_head;
@ -139,6 +146,7 @@ nvd_load()
if (!nvme_use_nvd)
return 0;
mtx_init(&nvd_lock, "nvd_lock", NULL, MTX_DEF);
TAILQ_INIT(&ctrlr_head);
TAILQ_INIT(&disk_head);
@ -152,25 +160,25 @@ static void
nvd_unload()
{
struct nvd_controller *ctrlr;
struct nvd_disk *disk;
struct nvd_disk *ndisk;
if (!nvme_use_nvd)
return;
while (!TAILQ_EMPTY(&ctrlr_head)) {
ctrlr = TAILQ_FIRST(&ctrlr_head);
mtx_lock(&nvd_lock);
while ((ctrlr = TAILQ_FIRST(&ctrlr_head)) != NULL) {
TAILQ_REMOVE(&ctrlr_head, ctrlr, tailq);
TAILQ_FOREACH(ndisk, &ctrlr->disk_head, ctrlr_tailq)
nvd_gone(ndisk);
while (!TAILQ_EMPTY(&ctrlr->disk_head))
msleep(&ctrlr->disk_head, &nvd_lock, 0, "nvd_unload",0);
free(ctrlr, M_NVD);
}
while (!TAILQ_EMPTY(&disk_head)) {
disk = TAILQ_FIRST(&disk_head);
TAILQ_REMOVE(&disk_head, disk, global_tailq);
destroy_geom_disk(disk);
free(disk, M_NVD);
}
mtx_unlock(&nvd_lock);
nvme_unregister_consumer(consumer_handle);
mtx_destroy(&nvd_lock);
}
static int
@ -220,6 +228,42 @@ nvd_strategy(struct bio *bp)
taskqueue_enqueue(ndisk->tq, &ndisk->bioqtask);
}
static void
nvd_gone(struct nvd_disk *ndisk)
{
struct bio *bp;
printf(NVD_STR"%u: detached\n", ndisk->unit);
mtx_lock(&ndisk->bioqlock);
disk_gone(ndisk->disk);
while ((bp = bioq_takefirst(&ndisk->bioq)) != NULL) {
if (__predict_false(bp->bio_flags & BIO_ORDERED))
atomic_add_int(&ndisk->ordered_in_flight, -1);
bp->bio_error = ENXIO;
bp->bio_flags |= BIO_ERROR;
bp->bio_resid = bp->bio_bcount;
biodone(bp);
}
mtx_unlock(&ndisk->bioqlock);
}
static void
nvd_gonecb(struct disk *dp)
{
struct nvd_disk *ndisk = (struct nvd_disk *)dp->d_drv1;
disk_destroy(ndisk->disk);
mtx_lock(&nvd_lock);
TAILQ_REMOVE(&disk_head, ndisk, global_tailq);
TAILQ_REMOVE(&ndisk->ctrlr->disk_head, ndisk, ctrlr_tailq);
if (TAILQ_EMPTY(&ndisk->ctrlr->disk_head))
wakeup(&ndisk->ctrlr->disk_head);
mtx_unlock(&nvd_lock);
taskqueue_free(ndisk->tq);
mtx_destroy(&ndisk->bioqlock);
free(ndisk, M_NVD);
}
static int
nvd_ioctl(struct disk *ndisk, u_long cmd, void *data, int fflag,
struct thread *td)
@ -304,7 +348,9 @@ nvd_new_controller(struct nvme_controller *ctrlr)
M_ZERO | M_WAITOK);
TAILQ_INIT(&nvd_ctrlr->disk_head);
mtx_lock(&nvd_lock);
TAILQ_INSERT_TAIL(&ctrlr_head, nvd_ctrlr, tailq);
mtx_unlock(&nvd_lock);
return (nvd_ctrlr);
}
@ -313,46 +359,61 @@ static void *
nvd_new_disk(struct nvme_namespace *ns, void *ctrlr_arg)
{
uint8_t descr[NVME_MODEL_NUMBER_LENGTH+1];
struct nvd_disk *ndisk;
struct nvd_disk *ndisk, *tnd;
struct disk *disk;
struct nvd_controller *ctrlr = ctrlr_arg;
int unit;
ndisk = malloc(sizeof(struct nvd_disk), M_NVD, M_ZERO | M_WAITOK);
ndisk->ctrlr = ctrlr;
ndisk->ns = ns;
ndisk->cur_depth = 0;
ndisk->ordered_in_flight = 0;
mtx_init(&ndisk->bioqlock, "nvd bioq lock", NULL, MTX_DEF);
bioq_init(&ndisk->bioq);
TASK_INIT(&ndisk->bioqtask, 0, nvd_bioq_process, ndisk);
disk = disk_alloc();
mtx_lock(&nvd_lock);
unit = 0;
TAILQ_FOREACH(tnd, &disk_head, global_tailq) {
if (tnd->unit > unit)
break;
unit = tnd->unit + 1;
}
ndisk->unit = unit;
if (tnd != NULL)
TAILQ_INSERT_BEFORE(tnd, ndisk, global_tailq);
else
TAILQ_INSERT_TAIL(&disk_head, ndisk, global_tailq);
TAILQ_INSERT_TAIL(&ctrlr->disk_head, ndisk, ctrlr_tailq);
mtx_unlock(&nvd_lock);
ndisk->tq = taskqueue_create("nvd_taskq", M_WAITOK,
taskqueue_thread_enqueue, &ndisk->tq);
taskqueue_start_threads(&ndisk->tq, 1, PI_DISK, "nvd taskq");
disk = ndisk->disk = disk_alloc();
disk->d_strategy = nvd_strategy;
disk->d_ioctl = nvd_ioctl;
disk->d_dump = nvd_dump;
disk->d_gone = nvd_gonecb;
disk->d_name = NVD_STR;
disk->d_unit = ndisk->unit;
disk->d_drv1 = ndisk;
disk->d_maxsize = nvme_ns_get_max_io_xfer_size(ns);
disk->d_sectorsize = nvme_ns_get_sector_size(ns);
disk->d_mediasize = (off_t)nvme_ns_get_size(ns);
disk->d_maxsize = nvme_ns_get_max_io_xfer_size(ns);
disk->d_delmaxsize = (off_t)nvme_ns_get_size(ns);
if (disk->d_delmaxsize > nvd_delete_max)
disk->d_delmaxsize = nvd_delete_max;
disk->d_stripesize = nvme_ns_get_stripesize(ns);
if (TAILQ_EMPTY(&disk_head))
disk->d_unit = 0;
else
disk->d_unit =
TAILQ_LAST(&disk_head, disk_list)->disk->d_unit + 1;
disk->d_flags = DISKFLAG_DIRECT_COMPLETION;
disk->d_flags = DISKFLAG_UNMAPPED_BIO | DISKFLAG_DIRECT_COMPLETION;
if (nvme_ns_get_flags(ns) & NVME_NS_DEALLOCATE_SUPPORTED)
disk->d_flags |= DISKFLAG_CANDELETE;
if (nvme_ns_get_flags(ns) & NVME_NS_FLUSH_SUPPORTED)
disk->d_flags |= DISKFLAG_CANFLUSHCACHE;
/* ifdef used here to ease porting to stable branches at a later point. */
#ifdef DISKFLAG_UNMAPPED_BIO
disk->d_flags |= DISKFLAG_UNMAPPED_BIO;
#endif
/*
* d_ident and d_descr are both far bigger than the length of either
* the serial or model number strings.
@ -365,22 +426,6 @@ nvd_new_disk(struct nvme_namespace *ns, void *ctrlr_arg)
disk->d_rotation_rate = DISK_RR_NON_ROTATING;
ndisk->ns = ns;
ndisk->disk = disk;
ndisk->cur_depth = 0;
ndisk->ordered_in_flight = 0;
mtx_init(&ndisk->bioqlock, "NVD bioq lock", NULL, MTX_DEF);
bioq_init(&ndisk->bioq);
TASK_INIT(&ndisk->bioqtask, 0, nvd_bioq_process, ndisk);
ndisk->tq = taskqueue_create("nvd_taskq", M_WAITOK,
taskqueue_thread_enqueue, &ndisk->tq);
taskqueue_start_threads(&ndisk->tq, 1, PI_DISK, "nvd taskq");
TAILQ_INSERT_TAIL(&disk_head, ndisk, global_tailq);
TAILQ_INSERT_TAIL(&ctrlr->disk_head, ndisk, ctrlr_tailq);
disk_create(disk, DISK_VERSION);
printf(NVD_STR"%u: <%s> NVMe namespace\n", disk->d_unit, descr);
@ -389,58 +434,22 @@ nvd_new_disk(struct nvme_namespace *ns, void *ctrlr_arg)
(uintmax_t)disk->d_mediasize / disk->d_sectorsize,
disk->d_sectorsize);
return (NULL);
}
static void
destroy_geom_disk(struct nvd_disk *ndisk)
{
struct bio *bp;
struct disk *disk;
uint32_t unit;
int cnt = 0;
disk = ndisk->disk;
unit = disk->d_unit;
taskqueue_free(ndisk->tq);
disk_destroy(ndisk->disk);
mtx_lock(&ndisk->bioqlock);
for (;;) {
bp = bioq_takefirst(&ndisk->bioq);
if (bp == NULL)
break;
bp->bio_error = EIO;
bp->bio_flags |= BIO_ERROR;
bp->bio_resid = bp->bio_bcount;
cnt++;
biodone(bp);
}
printf(NVD_STR"%u: lost device - %d outstanding\n", unit, cnt);
printf(NVD_STR"%u: removing device entry\n", unit);
mtx_unlock(&ndisk->bioqlock);
mtx_destroy(&ndisk->bioqlock);
return (ndisk);
}
static void
nvd_controller_fail(void *ctrlr_arg)
{
struct nvd_controller *ctrlr = ctrlr_arg;
struct nvd_disk *disk;
while (!TAILQ_EMPTY(&ctrlr->disk_head)) {
disk = TAILQ_FIRST(&ctrlr->disk_head);
TAILQ_REMOVE(&disk_head, disk, global_tailq);
TAILQ_REMOVE(&ctrlr->disk_head, disk, ctrlr_tailq);
destroy_geom_disk(disk);
free(disk, M_NVD);
}
struct nvd_disk *ndisk;
mtx_lock(&nvd_lock);
TAILQ_REMOVE(&ctrlr_head, ctrlr, tailq);
TAILQ_FOREACH(ndisk, &ctrlr->disk_head, ctrlr_tailq)
nvd_gone(ndisk);
while (!TAILQ_EMPTY(&ctrlr->disk_head))
msleep(&ctrlr->disk_head, &nvd_lock, 0, "nvd_fail", 0);
mtx_unlock(&nvd_lock);
free(ctrlr, M_NVD);
}