Merge branch 'akpm' (Andrew's patch-bomb)

Merge misc patches from Andrew Morton:
 "Incoming:

   - lots of misc stuff

   - backlight tree updates

   - lib/ updates

   - Oleg's percpu-rwsem changes

   - checkpatch

   - rtc

   - aoe

   - more checkpoint/restart support

  I still have a pile of MM stuff pending - Pekka should be merging
  later today after which that is good to go.  A number of other things
  are twiddling thumbs awaiting maintainer merges."

* emailed patches from Andrew Morton <[email protected]>: (180 commits)
  scatterlist: don't BUG when we can trivially return a proper error.
  docs: update documentation about /proc/<pid>/fdinfo/<fd> fanotify output
  fs, fanotify: add @mflags field to fanotify output
  docs: add documentation about /proc/<pid>/fdinfo/<fd> output
  fs, notify: add procfs fdinfo helper
  fs, exportfs: add exportfs_encode_inode_fh() helper
  fs, exportfs: escape nil dereference if no s_export_op present
  fs, epoll: add procfs fdinfo helper
  fs, eventfd: add procfs fdinfo helper
  procfs: add ability to plug in auxiliary fdinfo providers
  tools/testing/selftests/kcmp/kcmp_test.c: print reason for failure in kcmp_test
  breakpoint selftests: print failure status instead of cause make error
  kcmp selftests: print fail status instead of cause make error
  kcmp selftests: make run_tests fix
  mem-hotplug selftests: print failure status instead of cause make error
  cpu-hotplug selftests: print failure status instead of cause make error
  mqueue selftests: print failure status instead of cause make error
  vm selftests: print failure status instead of cause make error
  ubifs: use prandom_bytes
  mtd: nandsim: use prandom_bytes
  ...

drivers/block/aoe/aoe.h

@@ -1,5 +1,5 @@
 /* Copyright (c) 2012 Coraid, Inc.  See COPYING for GPL terms. */
-#define VERSION "50"
+#define VERSION "81"
 #define AOE_MAJOR 152
 #define DEVICE_NAME "aoe"
 
@@ -10,7 +10,7 @@
 #define AOE_PARTITIONS (16)
 #endif
 
-#define WHITESPACE " \t\v\f\n"
+#define WHITESPACE " \t\v\f\n,"
 
 enum {
 	AOECMD_ATA,
@@ -73,21 +73,29 @@ enum {
 	DEVFL_TKILL = (1<<1),	/* flag for timer to know when to kill self */
 	DEVFL_EXT = (1<<2),	/* device accepts lba48 commands */
 	DEVFL_GDALLOC = (1<<3),	/* need to alloc gendisk */
-	DEVFL_KICKME = (1<<4),	/* slow polling network card catch */
-	DEVFL_NEWSIZE = (1<<5),	/* need to update dev size in block layer */
+	DEVFL_GD_NOW = (1<<4),	/* allocating gendisk */
+	DEVFL_KICKME = (1<<5),	/* slow polling network card catch */
+	DEVFL_NEWSIZE = (1<<6),	/* need to update dev size in block layer */
+	DEVFL_FREEING = (1<<7),	/* set when device is being cleaned up */
+	DEVFL_FREED = (1<<8),	/* device has been cleaned up */
 };
 
 enum {
 	DEFAULTBCNT = 2 * 512,	/* 2 sectors */
 	MIN_BUFS = 16,
-	NTARGETS = 8,
+	NTARGETS = 4,
 	NAOEIFS = 8,
 	NSKBPOOLMAX = 256,
 	NFACTIVE = 61,
 
 	TIMERTICK = HZ / 10,
-	MINTIMER = HZ >> 2,
-	MAXTIMER = HZ << 1,
+	RTTSCALE = 8,
+	RTTDSCALE = 3,
+	RTTAVG_INIT = USEC_PER_SEC / 4 << RTTSCALE,
+	RTTDEV_INIT = RTTAVG_INIT / 4,
+
+	HARD_SCORN_SECS = 10,	/* try another remote port after this */
+	MAX_TAINT = 1000,	/* cap on aoetgt taint */
 };
 
 struct buf {
@@ -100,10 +108,17 @@ struct buf {
 	struct request *rq;
 };
 
+enum frame_flags {
+	FFL_PROBE = 1,
+};
+
 struct frame {
 	struct list_head head;
 	u32 tag;
+	struct timeval sent;	/* high-res time packet was sent */
+	u32 sent_jiffs;		/* low-res jiffies-based sent time */
 	ulong waited;
+	ulong waited_total;
 	struct aoetgt *t;		/* parent target I belong to */
 	sector_t lba;
 	struct sk_buff *skb;		/* command skb freed on module exit */
@@ -112,6 +127,7 @@ struct frame {
 	struct bio_vec *bv;
 	ulong bcnt;
 	ulong bv_off;
+	char flags;
 };
 
 struct aoeif {
@@ -122,28 +138,31 @@ struct aoeif {
 
 struct aoetgt {
 	unsigned char addr[6];
-	ushort nframes;
+	ushort nframes;		/* cap on frames to use */
 	struct aoedev *d;			/* parent device I belong to */
 	struct list_head ffree;			/* list of free frames */
 	struct aoeif ifs[NAOEIFS];
 	struct aoeif *ifp;	/* current aoeif in use */
-	ushort nout;
-	ushort maxout;
-	ulong falloc;
-	ulong lastwadj;		/* last window adjustment */
+	ushort nout;		/* number of AoE commands outstanding */
+	ushort maxout;		/* current value for max outstanding */
+	ushort next_cwnd;	/* incr maxout after decrementing to zero */
+	ushort ssthresh;	/* slow start threshold */
+	ulong falloc;		/* number of allocated frames */
+	int taint;		/* how much we want to avoid this aoetgt */
 	int minbcnt;
 	int wpkts, rpkts;
+	char nout_probes;
 };
 
 struct aoedev {
 	struct aoedev *next;
 	ulong sysminor;
 	ulong aoemajor;
+	u32 rttavg;		/* scaled AoE round trip time average */
+	u32 rttdev;		/* scaled round trip time mean deviation */
 	u16 aoeminor;
 	u16 flags;
 	u16 nopen;		/* (bd_openers isn't available without sleeping) */
-	u16 rttavg;		/* round trip average of requests/responses */
-	u16 mintimer;
 	u16 fw_ver;		/* version of blade's firmware */
 	u16 lasttag;		/* last tag sent */
 	u16 useme;
@@ -151,7 +170,7 @@ struct aoedev {
 	struct work_struct work;/* disk create work struct */
 	struct gendisk *gd;
 	struct request_queue *blkq;
-	struct hd_geometry geo; 
+	struct hd_geometry geo;
 	sector_t ssize;
 	struct timer_list timer;
 	spinlock_t lock;
@@ -164,11 +183,12 @@ struct aoedev {
 	} ip;
 	ulong maxbcnt;
 	struct list_head factive[NFACTIVE];	/* hash of active frames */
-	struct aoetgt *targets[NTARGETS];
+	struct list_head rexmitq; /* deferred retransmissions */
+	struct aoetgt **targets;
+	ulong ntargets;		/* number of allocated aoetgt pointers */
 	struct aoetgt **tgt;	/* target in use when working */
-	struct aoetgt *htgt;	/* target needing rexmit assistance */
-	ulong ntargets;
 	ulong kicked;
+	char ident[512];
 };
 
 /* kthread tracking */
@@ -195,6 +215,7 @@ void aoecmd_cfg(ushort aoemajor, unsigned char aoeminor);
 struct sk_buff *aoecmd_ata_rsp(struct sk_buff *);
 void aoecmd_cfg_rsp(struct sk_buff *);
 void aoecmd_sleepwork(struct work_struct *);
+void aoecmd_wreset(struct aoetgt *t);
 void aoecmd_cleanslate(struct aoedev *);
 void aoecmd_exit(void);
 int aoecmd_init(void);

drivers/block/aoe/aoeblk.c

@@ -16,11 +16,19 @@
 #include <linux/netdevice.h>
 #include <linux/mutex.h>
 #include <linux/export.h>
+#include <linux/moduleparam.h>
+#include <scsi/sg.h>
 #include "aoe.h"
 
 static DEFINE_MUTEX(aoeblk_mutex);
 static struct kmem_cache *buf_pool_cache;
 
+/* GPFS needs a larger value than the default. */
+static int aoe_maxsectors;
+module_param(aoe_maxsectors, int, 0644);
+MODULE_PARM_DESC(aoe_maxsectors,
+	"When nonzero, set the maximum number of sectors per I/O request");
+
 static ssize_t aoedisk_show_state(struct device *dev,
 				  struct device_attribute *attr, char *page)
 {
@@ -59,7 +67,7 @@ static ssize_t aoedisk_show_netif(struct device *dev,
 	nd = nds;
 	ne = nd + ARRAY_SIZE(nds);
 	t = d->targets;
-	te = t + NTARGETS;
+	te = t + d->ntargets;
 	for (; t < te && *t; t++) {
 		ifp = (*t)->ifs;
 		e = ifp + NAOEIFS;
@@ -91,6 +99,14 @@ static ssize_t aoedisk_show_fwver(struct device *dev,
 
 	return snprintf(page, PAGE_SIZE, "0x%04x\n", (unsigned int) d->fw_ver);
 }
+static ssize_t aoedisk_show_payload(struct device *dev,
+				    struct device_attribute *attr, char *page)
+{
+	struct gendisk *disk = dev_to_disk(dev);
+	struct aoedev *d = disk->private_data;
+
+	return snprintf(page, PAGE_SIZE, "%lu\n", d->maxbcnt);
+}
 
 static DEVICE_ATTR(state, S_IRUGO, aoedisk_show_state, NULL);
 static DEVICE_ATTR(mac, S_IRUGO, aoedisk_show_mac, NULL);
@@ -99,12 +115,14 @@ static struct device_attribute dev_attr_firmware_version = {
 	.attr = { .name = "firmware-version", .mode = S_IRUGO },
 	.show = aoedisk_show_fwver,
 };
+static DEVICE_ATTR(payload, S_IRUGO, aoedisk_show_payload, NULL);
 
 static struct attribute *aoe_attrs[] = {
 	&dev_attr_state.attr,
 	&dev_attr_mac.attr,
 	&dev_attr_netif.attr,
 	&dev_attr_firmware_version.attr,
+	&dev_attr_payload.attr,
 	NULL,
 };
 
@@ -129,9 +147,18 @@ aoeblk_open(struct block_device *bdev, fmode_t mode)
 	struct aoedev *d = bdev->bd_disk->private_data;
 	ulong flags;
 
+	if (!virt_addr_valid(d)) {
+		pr_crit("aoe: invalid device pointer in %s\n",
+			__func__);
+		WARN_ON(1);
+		return -ENODEV;
+	}
+	if (!(d->flags & DEVFL_UP) || d->flags & DEVFL_TKILL)
+		return -ENODEV;
+
 	mutex_lock(&aoeblk_mutex);
 	spin_lock_irqsave(&d->lock, flags);
-	if (d->flags & DEVFL_UP) {
+	if (d->flags & DEVFL_UP && !(d->flags & DEVFL_TKILL)) {
 		d->nopen++;
 		spin_unlock_irqrestore(&d->lock, flags);
 		mutex_unlock(&aoeblk_mutex);
@@ -195,9 +222,38 @@ aoeblk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 	return 0;
 }
 
+static int
+aoeblk_ioctl(struct block_device *bdev, fmode_t mode, uint cmd, ulong arg)
+{
+	struct aoedev *d;
+
+	if (!arg)
+		return -EINVAL;
+
+	d = bdev->bd_disk->private_data;
+	if ((d->flags & DEVFL_UP) == 0) {
+		pr_err("aoe: disk not up\n");
+		return -ENODEV;
+	}
+
+	if (cmd == HDIO_GET_IDENTITY) {
+		if (!copy_to_user((void __user *) arg, &d->ident,
+			sizeof(d->ident)))
+			return 0;
+		return -EFAULT;
+	}
+
+	/* udev calls scsi_id, which uses SG_IO, resulting in noise */
+	if (cmd != SG_IO)
+		pr_info("aoe: unknown ioctl 0x%x\n", cmd);
+
+	return -ENOTTY;
+}
+
 static const struct block_device_operations aoe_bdops = {
 	.open = aoeblk_open,
 	.release = aoeblk_release,
+	.ioctl = aoeblk_ioctl,
 	.getgeo = aoeblk_getgeo,
 	.owner = THIS_MODULE,
 };
@@ -212,6 +268,18 @@ aoeblk_gdalloc(void *vp)
 	struct request_queue *q;
 	enum { KB = 1024, MB = KB * KB, READ_AHEAD = 2 * MB, };
 	ulong flags;
+	int late = 0;
+
+	spin_lock_irqsave(&d->lock, flags);
+	if (d->flags & DEVFL_GDALLOC
+	&& !(d->flags & DEVFL_TKILL)
+	&& !(d->flags & DEVFL_GD_NOW))
+		d->flags |= DEVFL_GD_NOW;
+	else
+		late = 1;
+	spin_unlock_irqrestore(&d->lock, flags);
+	if (late)
+		return;
 
 	gd = alloc_disk(AOE_PARTITIONS);
 	if (gd == NULL) {
@@ -231,23 +299,24 @@ aoeblk_gdalloc(void *vp)
 	if (q == NULL) {
 		pr_err("aoe: cannot allocate block queue for %ld.%d\n",
 			d->aoemajor, d->aoeminor);
-		mempool_destroy(mp);
-		goto err_disk;
+		goto err_mempool;
 	}
 
-	d->blkq = blk_alloc_queue(GFP_KERNEL);
-	if (!d->blkq)
-		goto err_mempool;
-	d->blkq->backing_dev_info.name = "aoe";
-	if (bdi_init(&d->blkq->backing_dev_info))
-		goto err_blkq;
 	spin_lock_irqsave(&d->lock, flags);
-	blk_queue_max_hw_sectors(d->blkq, BLK_DEF_MAX_SECTORS);
+	WARN_ON(!(d->flags & DEVFL_GD_NOW));
+	WARN_ON(!(d->flags & DEVFL_GDALLOC));
+	WARN_ON(d->flags & DEVFL_TKILL);
+	WARN_ON(d->gd);
+	WARN_ON(d->flags & DEVFL_UP);
+	blk_queue_max_hw_sectors(q, BLK_DEF_MAX_SECTORS);
+	q->backing_dev_info.name = "aoe";
 	q->backing_dev_info.ra_pages = READ_AHEAD / PAGE_CACHE_SIZE;
 	d->bufpool = mp;
 	d->blkq = gd->queue = q;
 	q->queuedata = d;
 	d->gd = gd;
+	if (aoe_maxsectors)
+		blk_queue_max_hw_sectors(q, aoe_maxsectors);
 	gd->major = AOE_MAJOR;
 	gd->first_minor = d->sysminor;
 	gd->fops = &aoe_bdops;
@@ -263,18 +332,21 @@ aoeblk_gdalloc(void *vp)
 
 	add_disk(gd);
 	aoedisk_add_sysfs(d);
+
+	spin_lock_irqsave(&d->lock, flags);
+	WARN_ON(!(d->flags & DEVFL_GD_NOW));
+	d->flags &= ~DEVFL_GD_NOW;
+	spin_unlock_irqrestore(&d->lock, flags);
 	return;
 
-err_blkq:
-	blk_cleanup_queue(d->blkq);
-	d->blkq = NULL;
 err_mempool:
-	mempool_destroy(d->bufpool);
+	mempool_destroy(mp);
 err_disk:
 	put_disk(gd);
 err:
 	spin_lock_irqsave(&d->lock, flags);
-	d->flags &= ~DEVFL_GDALLOC;
+	d->flags &= ~DEVFL_GD_NOW;
+	schedule_work(&d->work);
 	spin_unlock_irqrestore(&d->lock, flags);
 }
 

drivers/block/aoe/aoechr.c

@@ -39,6 +39,11 @@ struct ErrMsg {
 };
 
 static DEFINE_MUTEX(aoechr_mutex);
+
+/* A ring buffer of error messages, to be read through
+ * "/dev/etherd/err".  When no messages are present,
+ * readers will block waiting for messages to appear.
+ */
 static struct ErrMsg emsgs[NMSG];
 static int emsgs_head_idx, emsgs_tail_idx;
 static struct completion emsgs_comp;
@@ -282,7 +287,7 @@ aoechr_init(void)
 	int n, i;
 
 	n = register_chrdev(AOE_MAJOR, "aoechr", &aoe_fops);
-	if (n < 0) { 
+	if (n < 0) {
 		printk(KERN_ERR "aoe: can't register char device\n");
 		return n;
 	}

drivers/block/aoe/aoedev.c

@@ -15,7 +15,6 @@
 #include "aoe.h"
 
 static void dummy_timer(ulong);
-static void aoedev_freedev(struct aoedev *);
 static void freetgt(struct aoedev *d, struct aoetgt *t);
 static void skbpoolfree(struct aoedev *d);
 
@@ -69,25 +68,34 @@ minor_get_static(ulong *sysminor, ulong aoemaj, int aoemin)
 		NPERSHELF = 16,
 	};
 
+	if (aoemin >= NPERSHELF) {
+		pr_err("aoe: %s %d slots per shelf\n",
+			"static minor device numbers support only",
+			NPERSHELF);
+		error = -1;
+		goto out;
+	}
+
 	n = aoemaj * NPERSHELF + aoemin;
-	if (aoemin >= NPERSHELF || n >= N_DEVS) {
+	if (n >= N_DEVS) {
 		pr_err("aoe: %s with e%ld.%d\n",
 			"cannot use static minor device numbers",
 			aoemaj, aoemin);
 		error = -1;
-	} else {
-		spin_lock_irqsave(&used_minors_lock, flags);
-		if (test_bit(n, used_minors)) {
-			pr_err("aoe: %s %lu\n",
-				"existing device already has static minor number",
-				n);
-			error = -1;
-		} else
-			set_bit(n, used_minors);
-		spin_unlock_irqrestore(&used_minors_lock, flags);
+		goto out;
 	}
 
-	*sysminor = n;
+	spin_lock_irqsave(&used_minors_lock, flags);
+	if (test_bit(n, used_minors)) {
+		pr_err("aoe: %s %lu\n",
+			"existing device already has static minor number",
+			n);
+		error = -1;
+	} else
+		set_bit(n, used_minors);
+	spin_unlock_irqrestore(&used_minors_lock, flags);
+	*sysminor = n * AOE_PARTITIONS;
+out:
 	return error;
 }
 
@@ -170,41 +178,50 @@ aoe_failip(struct aoedev *d)
 		aoe_end_request(d, rq, 0);
 }
 
+static void
+downdev_frame(struct list_head *pos)
+{
+	struct frame *f;
+
+	f = list_entry(pos, struct frame, head);
+	list_del(pos);
+	if (f->buf) {
+		f->buf->nframesout--;
+		aoe_failbuf(f->t->d, f->buf);
+	}
+	aoe_freetframe(f);
+}
+
 void
 aoedev_downdev(struct aoedev *d)
 {
 	struct aoetgt *t, **tt, **te;
-	struct frame *f;
 	struct list_head *head, *pos, *nx;
 	struct request *rq;
 	int i;
 
 	d->flags &= ~DEVFL_UP;
 
-	/* clean out active buffers */
+	/* clean out active and to-be-retransmitted buffers */
 	for (i = 0; i < NFACTIVE; i++) {
 		head = &d->factive[i];
-		list_for_each_safe(pos, nx, head) {
-			f = list_entry(pos, struct frame, head);
-			list_del(pos);
-			if (f->buf) {
-				f->buf->nframesout--;
-				aoe_failbuf(d, f->buf);
-			}
-			aoe_freetframe(f);
-		}
+		list_for_each_safe(pos, nx, head)
+			downdev_frame(pos);
 	}
+	head = &d->rexmitq;
+	list_for_each_safe(pos, nx, head)
+		downdev_frame(pos);
+
 	/* reset window dressings */
 	tt = d->targets;
-	te = tt + NTARGETS;
+	te = tt + d->ntargets;
 	for (; tt < te && (t = *tt); tt++) {
-		t->maxout = t->nframes;
+		aoecmd_wreset(t);
 		t->nout = 0;
 	}
 
 	/* clean out the in-process request (if any) */
 	aoe_failip(d);
-	d->htgt = NULL;
 
 	/* fast fail all pending I/O */
 	if (d->blkq) {
@@ -218,12 +235,48 @@ aoedev_downdev(struct aoedev *d)
 		set_capacity(d->gd, 0);
 }
 
+/* return whether the user asked for this particular
+ * device to be flushed
+ */
+static int
+user_req(char *s, size_t slen, struct aoedev *d)
+{
+	char *p;
+	size_t lim;
+
+	if (!d->gd)
+		return 0;
+	p = strrchr(d->gd->disk_name, '/');
+	if (!p)
+		p = d->gd->disk_name;
+	else
+		p += 1;
+	lim = sizeof(d->gd->disk_name);
+	lim -= p - d->gd->disk_name;
+	if (slen < lim)
+		lim = slen;
+
+	return !strncmp(s, p, lim);
+}
+
 static void
-aoedev_freedev(struct aoedev *d)
+freedev(struct aoedev *d)
 {
 	struct aoetgt **t, **e;
+	int freeing = 0;
+	unsigned long flags;
 
-	cancel_work_sync(&d->work);
+	spin_lock_irqsave(&d->lock, flags);
+	if (d->flags & DEVFL_TKILL
+	&& !(d->flags & DEVFL_FREEING)) {
+		d->flags |= DEVFL_FREEING;
+		freeing = 1;
+	}
+	spin_unlock_irqrestore(&d->lock, flags);
+	if (!freeing)
+		return;
+
+	del_timer_sync(&d->timer);
 	if (d->gd) {
 		aoedisk_rm_sysfs(d);
 		del_gendisk(d->gd);
@@ -231,61 +284,113 @@ aoedev_freedev(struct aoedev *d)
 		blk_cleanup_queue(d->blkq);
 	}
 	t = d->targets;
-	e = t + NTARGETS;
+	e = t + d->ntargets;
 	for (; t < e && *t; t++)
 		freetgt(d, *t);
 	if (d->bufpool)
 		mempool_destroy(d->bufpool);
 	skbpoolfree(d);
 	minor_free(d->sysminor);
-	kfree(d);
+
+	spin_lock_irqsave(&d->lock, flags);
+	d->flags |= DEVFL_FREED;
+	spin_unlock_irqrestore(&d->lock, flags);
 }
 
-int
-aoedev_flush(const char __user *str, size_t cnt)
+enum flush_parms {
+	NOT_EXITING = 0,
+	EXITING = 1,
+};
+
+static int
+flush(const char __user *str, size_t cnt, int exiting)
 {
 	ulong flags;
 	struct aoedev *d, **dd;
-	struct aoedev *rmd = NULL;
 	char buf[16];
 	int all = 0;
+	int specified = 0;	/* flush a specific device */
+	unsigned int skipflags;
 
-	if (cnt >= 3) {
+	skipflags = DEVFL_GDALLOC | DEVFL_NEWSIZE | DEVFL_TKILL;
+
+	if (!exiting && cnt >= 3) {
 		if (cnt > sizeof buf)
 			cnt = sizeof buf;
 		if (copy_from_user(buf, str, cnt))
 			return -EFAULT;
 		all = !strncmp(buf, "all", 3);
+		if (!all)
+			specified = 1;
 	}
 
+	flush_scheduled_work();
+	/* pass one: without sleeping, do aoedev_downdev */
 	spin_lock_irqsave(&devlist_lock, flags);
-	dd = &devlist;
-	while ((d = *dd)) {
+	for (d = devlist; d; d = d->next) {
 		spin_lock(&d->lock);
-		if ((!all && (d->flags & DEVFL_UP))
-		|| (d->flags & (DEVFL_GDALLOC|DEVFL_NEWSIZE))
+		if (exiting) {
+			/* unconditionally take each device down */
+		} else if (specified) {
+			if (!user_req(buf, cnt, d))
+				goto cont;
+		} else if ((!all && (d->flags & DEVFL_UP))
+		|| d->flags & skipflags
 		|| d->nopen
-		|| d->ref) {
-			spin_unlock(&d->lock);
-			dd = &d->next;
-			continue;
-		}
-		*dd = d->next;
+		|| d->ref)
+			goto cont;
+
 		aoedev_downdev(d);
 		d->flags |= DEVFL_TKILL;
+cont:
 		spin_unlock(&d->lock);
-		d->next = rmd;
-		rmd = d;
 	}
 	spin_unlock_irqrestore(&devlist_lock, flags);
-	while ((d = rmd)) {
-		rmd = d->next;
-		del_timer_sync(&d->timer);
-		aoedev_freedev(d);	/* must be able to sleep */
+
+	/* pass two: call freedev, which might sleep,
+	 * for aoedevs marked with DEVFL_TKILL
+	 */
+restart:
+	spin_lock_irqsave(&devlist_lock, flags);
+	for (d = devlist; d; d = d->next) {
+		spin_lock(&d->lock);
+		if (d->flags & DEVFL_TKILL
+		&& !(d->flags & DEVFL_FREEING)) {
+			spin_unlock(&d->lock);
+			spin_unlock_irqrestore(&devlist_lock, flags);
+			freedev(d);
+			goto restart;
+		}
+		spin_unlock(&d->lock);
 	}
+
+	/* pass three: remove aoedevs marked with DEVFL_FREED */
+	for (dd = &devlist, d = *dd; d; d = *dd) {
+		struct aoedev *doomed = NULL;
+
+		spin_lock(&d->lock);
+		if (d->flags & DEVFL_FREED) {
+			*dd = d->next;
+			doomed = d;
+		} else {
+			dd = &d->next;
+		}
+		spin_unlock(&d->lock);
+		if (doomed)
+			kfree(doomed->targets);
+		kfree(doomed);
+	}
+	spin_unlock_irqrestore(&devlist_lock, flags);
+
 	return 0;
 }
 
+int
+aoedev_flush(const char __user *str, size_t cnt)
+{
+	return flush(str, cnt, NOT_EXITING);
+}
+
 /* This has been confirmed to occur once with Tms=3*1000 due to the
  * driver changing link and not processing its transmit ring.  The
  * problem is hard enough to solve by returning an error that I'm
@@ -332,13 +437,20 @@ aoedev_by_aoeaddr(ulong maj, int min, int do_alloc)
 	struct aoedev *d;
 	int i;
 	ulong flags;
-	ulong sysminor;
+	ulong sysminor = 0;
 
 	spin_lock_irqsave(&devlist_lock, flags);
 
 	for (d=devlist; d; d=d->next)
 		if (d->aoemajor == maj && d->aoeminor == min) {
+			spin_lock(&d->lock);
+			if (d->flags & DEVFL_TKILL) {
+				spin_unlock(&d->lock);
+				d = NULL;
+				goto out;
+			}
 			d->ref++;
+			spin_unlock(&d->lock);
 			break;
 		}
 	if (d || !do_alloc || minor_get(&sysminor, maj, min) < 0)
@@ -346,6 +458,13 @@ aoedev_by_aoeaddr(ulong maj, int min, int do_alloc)
 	d = kcalloc(1, sizeof *d, GFP_ATOMIC);
 	if (!d)
 		goto out;
+	d->targets = kcalloc(NTARGETS, sizeof(*d->targets), GFP_ATOMIC);
+	if (!d->targets) {
+		kfree(d);
+		d = NULL;
+		goto out;
+	}
+	d->ntargets = NTARGETS;
 	INIT_WORK(&d->work, aoecmd_sleepwork);
 	spin_lock_init(&d->lock);
 	skb_queue_head_init(&d->skbpool);
@@ -359,10 +478,12 @@ aoedev_by_aoeaddr(ulong maj, int min, int do_alloc)
 	d->ref = 1;
 	for (i = 0; i < NFACTIVE; i++)
 		INIT_LIST_HEAD(&d->factive[i]);
+	INIT_LIST_HEAD(&d->rexmitq);
 	d->sysminor = sysminor;
 	d->aoemajor = maj;
 	d->aoeminor = min;
-	d->mintimer = MINTIMER;
+	d->rttavg = RTTAVG_INIT;
+	d->rttdev = RTTDEV_INIT;
 	d->next = devlist;
 	devlist = d;
  out:
@@ -396,21 +517,9 @@ freetgt(struct aoedev *d, struct aoetgt *t)
 void
 aoedev_exit(void)
 {
-	struct aoedev *d;
-	ulong flags;
-
+	flush_scheduled_work();
 	aoe_flush_iocq();
-	while ((d = devlist)) {
-		devlist = d->next;
-
-		spin_lock_irqsave(&d->lock, flags);
-		aoedev_downdev(d);
-		d->flags |= DEVFL_TKILL;
-		spin_unlock_irqrestore(&d->lock, flags);
-
-		del_timer_sync(&d->timer);
-		aoedev_freedev(d);
-	}
+	flush(NULL, 0, EXITING);
 }
 
 int __init

drivers/block/aoe/aoemain.c

@@ -105,7 +105,7 @@ aoe_init(void)
 	aoechr_exit();
  chr_fail:
 	aoedev_exit();
-	
+
 	printk(KERN_INFO "aoe: initialisation failure.\n");
 	return ret;
 }

drivers/block/aoe/aoenet.c

@@ -31,7 +31,7 @@ enum {
 
 static char aoe_iflist[IFLISTSZ];
 module_param_string(aoe_iflist, aoe_iflist, IFLISTSZ, 0600);
-MODULE_PARM_DESC(aoe_iflist, "aoe_iflist=\"dev1 [dev2 ...]\"");
+MODULE_PARM_DESC(aoe_iflist, "aoe_iflist=dev1[,dev2...]");
 
 static wait_queue_head_t txwq;
 static struct ktstate kts;
@@ -52,13 +52,18 @@ static struct sk_buff_head skbtxq;
 
 /* enters with txlock held */
 static int
-tx(void)
+tx(void) __must_hold(&txlock)
 {
 	struct sk_buff *skb;
+	struct net_device *ifp;
 
 	while ((skb = skb_dequeue(&skbtxq))) {
 		spin_unlock_irq(&txlock);
-		dev_queue_xmit(skb);
+		ifp = skb->dev;
+		if (dev_queue_xmit(skb) == NET_XMIT_DROP && net_ratelimit())
+			pr_warn("aoe: packet could not be sent on %s.  %s\n",
+				ifp ? ifp->name : "netif",
+				"consider increasing tx_queue_len");
 		spin_lock_irq(&txlock);
 	}
 	return 0;
@@ -119,8 +124,8 @@ aoenet_xmit(struct sk_buff_head *queue)
 	}
 }
 
-/* 
- * (1) len doesn't include the header by default.  I want this. 
+/*
+ * (1) len doesn't include the header by default.  I want this.
  */
 static int
 aoenet_rcv(struct sk_buff *skb, struct net_device *ifp, struct packet_type *pt, struct net_device *orig_dev)