Merge git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/aoe-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/aoe-2.6:
  aoe: fix sysfs_create_file warnings
  aoe: revert printk macros
  aoe: update driver version
  aoe: remove sysfs comment
  aoe: use bio->bi_idx
  aoe: module parameter for device timeout
  aoe: zero copy write 2 of 2
  aoe: improve retransmission heuristics
  aoe: jumbo frame support 2 of 2
  aoe: clean up printks via macros
  aoe: jumbo frame support 1 of 2
  aoe: zero copy write 1 of 2
  aoe: remove unused NARGS enum
  aoe: update copyright date
  aoe: eliminate isbusy message

7 files changed, 287 insertions, 179 deletions

diff --git a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
index 6eebcb7be97..6d111228cfa 100644
--- a/drivers/block/aoe/aoe.h
+++ b/drivers/block/aoe/aoe.h
@@ -1,5 +1,5 @@
-/* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
-#define VERSION "22"
+/* Copyright (c) 2006 Coraid, Inc.  See COPYING for GPL terms. */
+#define VERSION "32"
 #define AOE_MAJOR 152
 #define DEVICE_NAME "aoe"
 
@@ -65,7 +65,7 @@ struct aoe_atahdr {
 struct aoe_cfghdr {
 	__be16 bufcnt;
 	__be16 fwver;
-	unsigned char res;
+	unsigned char scnt;
 	unsigned char aoeccmd;
 	unsigned char cslen[2];
 };
@@ -78,12 +78,14 @@ enum {
 	DEVFL_GDALLOC = (1<<4),	/* need to alloc gendisk */
 	DEVFL_PAUSE = (1<<5),
 	DEVFL_NEWSIZE = (1<<6),	/* need to update dev size in block layer */
+	DEVFL_MAXBCNT = (1<<7), /* d->maxbcnt is not changeable */
+	DEVFL_KICKME = (1<<8),
 
 	BUFFL_FAIL = 1,
 };
 
 enum {
-	MAXATADATA = 1024,
+	DEFAULTBCNT = 2 * 512,	/* 2 sectors */
 	NPERSHELF = 16,		/* number of slots per shelf address */
 	FREETAG = -1,
 	MIN_BUFS = 8,
@@ -107,11 +109,9 @@ struct frame {
 	ulong waited;
 	struct buf *buf;
 	char *bufaddr;
-	int writedatalen;
-	int ndata;
-
-	/* largest possible */
-	unsigned char data[sizeof(struct aoe_hdr) + sizeof(struct aoe_atahdr)];
+	ulong bcnt;
+	sector_t lba;
+	struct sk_buff *skb;
 };
 
 struct aoedev {
@@ -121,9 +121,12 @@ struct aoedev {
 	ulong sysminor;
 	ulong aoemajor;
 	ulong aoeminor;
-	ulong nopen;		/* (bd_openers isn't available without sleeping) */
-	ulong rttavg;		/* round trip average of requests/responses */
+	u16 nopen;		/* (bd_openers isn't available without sleeping) */
+	u16 lasttag;		/* last tag sent */
+	u16 rttavg;		/* round trip average of requests/responses */
+	u16 mintimer;
 	u16 fw_ver;		/* version of blade's firmware */
+	u16 maxbcnt;
 	struct work_struct work;/* disk create work struct */
 	struct gendisk *gd;
 	request_queue_t blkq;
@@ -137,8 +140,8 @@ struct aoedev {
 	mempool_t *bufpool;	/* for deadlock-free Buf allocation */
 	struct list_head bufq;	/* queue of bios to work on */
 	struct buf *inprocess;	/* the one we're currently working on */
-	ulong lasttag;		/* last tag sent */
-	ulong nframes;		/* number of frames below */
+	ushort lostjumbo;
+	ushort nframes;		/* number of frames below */
 	struct frame *frames;
 };
 
@@ -157,6 +160,7 @@ void aoecmd_cfg(ushort aoemajor, unsigned char aoeminor);
 void aoecmd_ata_rsp(struct sk_buff *);
 void aoecmd_cfg_rsp(struct sk_buff *);
 void aoecmd_sleepwork(void *vp);
+struct sk_buff *new_skb(ulong);
 
 int aoedev_init(void);
 void aoedev_exit(void);
diff --git a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c
index 393b86a3dbf..d433f27e0ce 100644
--- a/drivers/block/aoe/aoeblk.c
+++ b/drivers/block/aoe/aoeblk.c
@@ -1,4 +1,4 @@
-/* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
+/* Copyright (c) 2006 Coraid, Inc.  See COPYING for GPL terms. */
 /*
  * aoeblk.c
  * block device routines
@@ -14,7 +14,6 @@
 
 static kmem_cache_t *buf_pool_cache;
 
-/* add attributes for our block devices in sysfs */
 static ssize_t aoedisk_show_state(struct gendisk * disk, char *page)
 {
 	struct aoedev *d = disk->private_data;
@@ -64,21 +63,26 @@ static struct disk_attribute disk_attr_fwver = {
 	.show = aoedisk_show_fwver
 };
 
-static void
+static struct attribute *aoe_attrs[] = {
+	&disk_attr_state.attr,
+	&disk_attr_mac.attr,
+	&disk_attr_netif.attr,
+	&disk_attr_fwver.attr,
+};
+
+static const struct attribute_group attr_group = {
+	.attrs = aoe_attrs,
+};
+
+static int
 aoedisk_add_sysfs(struct aoedev *d)
 {
-	sysfs_create_file(&d->gd->kobj, &disk_attr_state.attr);
-	sysfs_create_file(&d->gd->kobj, &disk_attr_mac.attr);
-	sysfs_create_file(&d->gd->kobj, &disk_attr_netif.attr);
-	sysfs_create_file(&d->gd->kobj, &disk_attr_fwver.attr);
+	return sysfs_create_group(&d->gd->kobj, &attr_group);
 }
 void
 aoedisk_rm_sysfs(struct aoedev *d)
 {
-	sysfs_remove_link(&d->gd->kobj, "state");
-	sysfs_remove_link(&d->gd->kobj, "mac");
-	sysfs_remove_link(&d->gd->kobj, "netif");
-	sysfs_remove_link(&d->gd->kobj, "firmware-version");
+	sysfs_remove_group(&d->gd->kobj, &attr_group);
 }
 
 static int
@@ -132,8 +136,7 @@ aoeblk_make_request(request_queue_t *q, struct bio *bio)
 	d = bio->bi_bdev->bd_disk->private_data;
 	buf = mempool_alloc(d->bufpool, GFP_NOIO);
 	if (buf == NULL) {
-		printk(KERN_INFO "aoe: aoeblk_make_request: buf allocation "
-			"failure\n");
+		printk(KERN_INFO "aoe: buf allocation failure\n");
 		bio_endio(bio, bio->bi_size, -ENOMEM);
 		return 0;
 	}
@@ -143,14 +146,15 @@ aoeblk_make_request(request_queue_t *q, struct bio *bio)
 	buf->bio = bio;
 	buf->resid = bio->bi_size;
 	buf->sector = bio->bi_sector;
-	buf->bv = buf->bio->bi_io_vec;
+	buf->bv = &bio->bi_io_vec[bio->bi_idx];
+	WARN_ON(buf->bv->bv_len == 0);
 	buf->bv_resid = buf->bv->bv_len;
 	buf->bufaddr = page_address(buf->bv->bv_page) + buf->bv->bv_offset;
 
 	spin_lock_irqsave(&d->lock, flags);
 
 	if ((d->flags & DEVFL_UP) == 0) {
-		printk(KERN_INFO "aoe: aoeblk_make_request: device %ld.%ld is not up\n",
+		printk(KERN_INFO "aoe: device %ld.%ld is not up\n",
 			d->aoemajor, d->aoeminor);
 		spin_unlock_irqrestore(&d->lock, flags);
 		mempool_free(buf, d->bufpool);
@@ -176,7 +180,7 @@ aoeblk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 	struct aoedev *d = bdev->bd_disk->private_data;
 
 	if ((d->flags & DEVFL_UP) == 0) {
-		printk(KERN_ERR "aoe: aoeblk_ioctl: disk not up\n");
+		printk(KERN_ERR "aoe: disk not up\n");
 		return -ENODEV;
 	}
 
@@ -203,8 +207,8 @@ aoeblk_gdalloc(void *vp)
 
 	gd = alloc_disk(AOE_PARTITIONS);
 	if (gd == NULL) {
-		printk(KERN_ERR "aoe: aoeblk_gdalloc: cannot allocate disk "
-			"structure for %ld.%ld\n", d->aoemajor, d->aoeminor);
+		printk(KERN_ERR "aoe: cannot allocate disk structure for %ld.%ld\n",
+			d->aoemajor, d->aoeminor);
 		spin_lock_irqsave(&d->lock, flags);
 		d->flags &= ~DEVFL_GDALLOC;
 		spin_unlock_irqrestore(&d->lock, flags);
@@ -213,8 +217,8 @@ aoeblk_gdalloc(void *vp)
 
 	d->bufpool = mempool_create_slab_pool(MIN_BUFS, buf_pool_cache);
 	if (d->bufpool == NULL) {
-		printk(KERN_ERR "aoe: aoeblk_gdalloc: cannot allocate bufpool "
-			"for %ld.%ld\n", d->aoemajor, d->aoeminor);
+		printk(KERN_ERR "aoe: cannot allocate bufpool for %ld.%ld\n",
+			d->aoemajor, d->aoeminor);
 		put_disk(gd);
 		spin_lock_irqsave(&d->lock, flags);
 		d->flags &= ~DEVFL_GDALLOC;
diff --git a/drivers/block/aoe/aoechr.c b/drivers/block/aoe/aoechr.c
index 1bc1cf9603f..e22b4c9520a 100644
--- a/drivers/block/aoe/aoechr.c
+++ b/drivers/block/aoe/aoechr.c
@@ -1,4 +1,4 @@
-/* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
+/* Copyright (c) 2006 Coraid, Inc.  See COPYING for GPL terms. */
 /*
  * aoechr.c
  * AoE character device driver
@@ -15,7 +15,6 @@ enum {
 	MINOR_INTERFACES,
 	MINOR_REVALIDATE,
 	MSGSZ = 2048,
-	NARGS = 10,
 	NMSG = 100,		/* message backlog to retain */
 };
 
@@ -56,9 +55,8 @@ static int
 interfaces(const char __user *str, size_t size)
 {
 	if (set_aoe_iflist(str, size)) {
-		printk(KERN_CRIT
-		       "%s: could not set interface list: %s\n",
-		       __FUNCTION__, "too many interfaces");
+		printk(KERN_ERR
+			"aoe: could not set interface list: too many interfaces\n");
 		return -EINVAL;
 	}
 	return 0;
@@ -81,8 +79,7 @@ revalidate(const char __user *str, size_t size)
 	/* should be e%d.%d format */
 	n = sscanf(buf, "e%d.%d", &major, &minor);
 	if (n != 2) {
-		printk(KERN_ERR "aoe: %s: invalid device specification\n",
-			__FUNCTION__);
+		printk(KERN_ERR "aoe: invalid device specification\n");
 		return -EINVAL;
 	}
 	d = aoedev_by_aoeaddr(major, minor);
@@ -90,6 +87,7 @@ revalidate(const char __user *str, size_t size)
 		return -EINVAL;
 
 	spin_lock_irqsave(&d->lock, flags);
+	d->flags &= ~DEVFL_MAXBCNT;
 	d->flags |= DEVFL_PAUSE;
 	spin_unlock_irqrestore(&d->lock, flags);
 	aoecmd_cfg(major, minor);
@@ -116,7 +114,7 @@ bail:		spin_unlock_irqrestore(&emsgs_lock, flags);
 
 	mp = kmalloc(n, GFP_ATOMIC);
 	if (mp == NULL) {
-		printk(KERN_CRIT "aoe: aoechr_error: allocation failure, len=%ld\n", n);
+		printk(KERN_ERR "aoe: allocation failure, len=%ld\n", n);
 		goto bail;
 	}
 
@@ -141,7 +139,7 @@ aoechr_write(struct file *filp, const char __user *buf, size_t cnt, loff_t *offp
 
 	switch ((unsigned long) filp->private_data) {
 	default:
-		printk(KERN_INFO "aoe: aoechr_write: can't write to that file.\n");
+		printk(KERN_INFO "aoe: can't write to that file.\n");
 		break;
 	case MINOR_DISCOVER:
 		ret = discover();
@@ -250,7 +248,7 @@ aoechr_init(void)
 
 	n = register_chrdev(AOE_MAJOR, "aoechr", &aoe_fops);
 	if (n < 0) { 
-		printk(KERN_ERR "aoe: aoechr_init: can't register char device\n");
+		printk(KERN_ERR "aoe: can't register char device\n");
 		return n;
 	}
 	sema_init(&emsgs_sema, 0);
diff --git a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
index 39da28d344f..8a13b1af8ba 100644
--- a/drivers/block/aoe/aoecmd.c
+++ b/drivers/block/aoe/aoecmd.c
@@ -1,4 +1,4 @@
-/* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
+/* Copyright (c) 2006 Coraid, Inc.  See COPYING for GPL terms. */
 /*
  * aoecmd.c
  * Filesystem request handling methods
@@ -15,17 +15,19 @@
 #define TIMERTICK (HZ / 10)
 #define MINTIMER (2 * TIMERTICK)
 #define MAXTIMER (HZ << 1)
-#define MAXWAIT (60 * 3)	/* After MAXWAIT seconds, give up and fail dev */
 
-static struct sk_buff *
-new_skb(struct net_device *if_dev, ulong len)
+static int aoe_deadsecs = 60 * 3;
+module_param(aoe_deadsecs, int, 0644);
+MODULE_PARM_DESC(aoe_deadsecs, "After aoe_deadsecs seconds, give up and fail dev.");
+
+struct sk_buff *
+new_skb(ulong len)
 {
 	struct sk_buff *skb;
 
 	skb = alloc_skb(len, GFP_ATOMIC);
 	if (skb) {
 		skb->nh.raw = skb->mac.raw = skb->data;
-		skb->dev = if_dev;
 		skb->protocol = __constant_htons(ETH_P_AOE);
 		skb->priority = 0;
 		skb_put(skb, len);
@@ -40,29 +42,6 @@ new_skb(struct net_device *if_dev, ulong len)
 	return skb;
 }
 
-static struct sk_buff *
-skb_prepare(struct aoedev *d, struct frame *f)
-{
-	struct sk_buff *skb;
-	char *p;
-
-	skb = new_skb(d->ifp, f->ndata + f->writedatalen);
-	if (!skb) {
-		printk(KERN_INFO "aoe: skb_prepare: failure to allocate skb\n");
-		return NULL;
-	}
-
-	p = skb->mac.raw;
-	memcpy(p, f->data, f->ndata);
-
-	if (f->writedatalen) {
-		p += sizeof(struct aoe_hdr) + sizeof(struct aoe_atahdr);
-		memcpy(p, f->bufaddr, f->writedatalen);
-	}
-
-	return skb;
-}
-
 static struct frame *
 getframe(struct aoedev *d, int tag)
 {
@@ -107,6 +86,17 @@ aoehdr_atainit(struct aoedev *d, struct aoe_hdr *h)
 	return host_tag;
 }
 
+static inline void
+put_lba(struct aoe_atahdr *ah, sector_t lba)
+{
+	ah->lba0 = lba;
+	ah->lba1 = lba >>= 8;
+	ah->lba2 = lba >>= 8;
+	ah->lba3 = lba >>= 8;
+	ah->lba4 = lba >>= 8;
+	ah->lba5 = lba >>= 8;
+}
+
 static void
 aoecmd_ata_rw(struct aoedev *d, struct frame *f)
 {
@@ -125,29 +115,27 @@ aoecmd_ata_rw(struct aoedev *d, struct frame *f)
 
 	sector = buf->sector;
 	bcnt = buf->bv_resid;
-	if (bcnt > MAXATADATA)
-		bcnt = MAXATADATA;
+	if (bcnt > d->maxbcnt)
+		bcnt = d->maxbcnt;
 
 	/* initialize the headers & frame */
-	h = (struct aoe_hdr *) f->data;
+	skb = f->skb;
+	h = (struct aoe_hdr *) skb->mac.raw;
 	ah = (struct aoe_atahdr *) (h+1);
-	f->ndata = sizeof *h + sizeof *ah;
-	memset(h, 0, f->ndata);
+	skb->len = sizeof *h + sizeof *ah;
+	memset(h, 0, ETH_ZLEN);
 	f->tag = aoehdr_atainit(d, h);
 	f->waited = 0;
 	f->buf = buf;
 	f->bufaddr = buf->bufaddr;
+	f->bcnt = bcnt;
+	f->lba = sector;
 
 	/* set up ata header */
 	ah->scnt = bcnt >> 9;
-	ah->lba0 = sector;
-	ah->lba1 = sector >>= 8;
-	ah->lba2 = sector >>= 8;
-	ah->lba3 = sector >>= 8;
+	put_lba(ah, sector);
 	if (d->flags & DEVFL_EXT) {
 		ah->aflags |= AOEAFL_EXT;
-		ah->lba4 = sector >>= 8;
-		ah->lba5 = sector >>= 8;
 	} else {
 		extbit = 0;
 		ah->lba3 &= 0x0f;
@@ -155,11 +143,14 @@ aoecmd_ata_rw(struct aoedev *d, struct frame *f)
 	}
 
 	if (bio_data_dir(buf->bio) == WRITE) {
+		skb_fill_page_desc(skb, 0, virt_to_page(f->bufaddr),
+			offset_in_page(f->bufaddr), bcnt);
 		ah->aflags |= AOEAFL_WRITE;
-		f->writedatalen = bcnt;
+		skb->len += bcnt;
+		skb->data_len = bcnt;
 	} else {
+		skb->len = ETH_ZLEN;
 		writebit = 0;
-		f->writedatalen = 0;
 	}
 
 	ah->cmdstat = WIN_READ | writebit | extbit;
@@ -168,26 +159,27 @@ aoecmd_ata_rw(struct aoedev *d, struct frame *f)
 	buf->nframesout += 1;
 	buf->bufaddr += bcnt;
 	buf->bv_resid -= bcnt;
-/* printk(KERN_INFO "aoe: bv_resid=%ld\n", buf->bv_resid); */
+/* printk(KERN_DEBUG "aoe: bv_resid=%ld\n", buf->bv_resid); */
 	buf->resid -= bcnt;
 	buf->sector += bcnt >> 9;
 	if (buf->resid == 0) {
 		d->inprocess = NULL;
 	} else if (buf->bv_resid == 0) {
 		buf->bv++;
+		WARN_ON(buf->bv->bv_len == 0);
 		buf->bv_resid = buf->bv->bv_len;
 		buf->bufaddr = page_address(buf->bv->bv_page) + buf->bv->bv_offset;
 	}
 
-	skb = skb_prepare(d, f);
-	if (skb) {
-		skb->next = NULL;
-		if (d->sendq_hd)
-			d->sendq_tl->next = skb;
-		else
-			d->sendq_hd = skb;
-		d->sendq_tl = skb;
-	}
+	skb->dev = d->ifp;
+	skb = skb_clone(skb, GFP_ATOMIC);
+	if (skb == NULL)
+		return;
+	if (d->sendq_hd)
+		d->sendq_tl->next = skb;
+	else
+		d->sendq_hd = skb;
+	d->sendq_tl = skb;
 }
 
 /* some callers cannot sleep, and they can call this function,
@@ -209,11 +201,12 @@ aoecmd_cfg_pkts(ushort aoemajor, unsigned char aoeminor, struct sk_buff **tail)
 		if (!is_aoe_netif(ifp))
 			continue;
 
-		skb = new_skb(ifp, sizeof *h + sizeof *ch);
+		skb = new_skb(sizeof *h + sizeof *ch);
 		if (skb == NULL) {
-			printk(KERN_INFO "aoe: aoecmd_cfg: skb alloc failure\n");
+			printk(KERN_INFO "aoe: skb alloc failure\n");
 			continue;
 		}
+		skb->dev = ifp;
 		if (sl_tail == NULL)
 			sl_tail = skb;
 		h = (struct aoe_hdr *) skb->mac.raw;
@@ -237,6 +230,29 @@ aoecmd_cfg_pkts(ushort aoemajor, unsigned char aoeminor, struct sk_buff **tail)
 	return sl;
 }
 
+static struct frame *
+freeframe(struct aoedev *d)
+{
+	struct frame *f, *e;
+	int n = 0;
+
+	f = d->frames;
+	e = f + d->nframes;
+	for (; f<e; f++) {
+		if (f->tag != FREETAG)
+			continue;
+		if (atomic_read(&skb_shinfo(f->skb)->dataref) == 1) {
+			skb_shinfo(f->skb)->nr_frags = f->skb->data_len = 0;
+			return f;
+		}
+		n++;
+	}
+	if (n == d->nframes)	/* wait for network layer */
+		d->flags |= DEVFL_KICKME;
+
+	return NULL;
+}
+
 /* enters with d->lock held */
 void
 aoecmd_work(struct aoedev *d)
@@ -252,7 +268,7 @@ aoecmd_work(struct aoedev *d)
 	}
 
 loop:
-	f = getframe(d, FREETAG);
+	f = freeframe(d);
 	if (f == NULL)
 		return;
 	if (d->inprocess == NULL) {
@@ -260,7 +276,7 @@ loop:
 			return;
 		buf = container_of(d->bufq.next, struct buf, bufs);
 		list_del(d->bufq.next);
-/*printk(KERN_INFO "aoecmd_work: bi_size=%ld\n", buf->bio->bi_size); */
+/*printk(KERN_DEBUG "aoe: bi_size=%ld\n", buf->bio->bi_size); */
 		d->inprocess = buf;
 	}
 	aoecmd_ata_rw(d, f);
@@ -272,6 +288,7 @@ rexmit(struct aoedev *d, struct frame *f)
 {
 	struct sk_buff *skb;
 	struct aoe_hdr *h;
+	struct aoe_atahdr *ah;
 	char buf[128];
 	u32 n;
 
@@ -283,21 +300,41 @@ rexmit(struct aoedev *d, struct frame *f)
 		d->aoemajor, d->aoeminor, f->tag, jiffies, n);
 	aoechr_error(buf);
 
-	h = (struct aoe_hdr *) f->data;
+	skb = f->skb;
+	h = (struct aoe_hdr *) skb->mac.raw;
+	ah = (struct aoe_atahdr *) (h+1);
 	f->tag = n;
 	h->tag = cpu_to_be32(n);
 	memcpy(h->dst, d->addr, sizeof h->dst);
 	memcpy(h->src, d->ifp->dev_addr, sizeof h->src);
 
-	skb = skb_prepare(d, f);
-	if (skb) {
-		skb->next = NULL;
-		if (d->sendq_hd)
-			d->sendq_tl->next = skb;
-		else
-			d->sendq_hd = skb;
-		d->sendq_tl = skb;
+	n = DEFAULTBCNT / 512;
+	if (ah->scnt > n) {
+		ah->scnt = n;
+		if (ah->aflags & AOEAFL_WRITE) {
+			skb_fill_page_desc(skb, 0, virt_to_page(f->bufaddr),
+				offset_in_page(f->bufaddr), DEFAULTBCNT);
+			skb->len = sizeof *h + sizeof *ah + DEFAULTBCNT;
+			skb->data_len = DEFAULTBCNT;
+		}
+		if (++d->lostjumbo > (d->nframes << 1))
+		if (d->maxbcnt != DEFAULTBCNT) {
+			printk(KERN_INFO "aoe: e%ld.%ld: too many lost jumbo on %s - using 1KB frames.\n",
+				d->aoemajor, d->aoeminor, d->ifp->name);
+			d->maxbcnt = DEFAULTBCNT;
+			d->flags |= DEVFL_MAXBCNT;
+		}
 	}
+
+	skb->dev = d->ifp;
+	skb = skb_clone(skb, GFP_ATOMIC);
+	if (skb == NULL)
+		return;
+	if (d->sendq_hd)
+		d->sendq_tl->next = skb;
+	else
+		d->sendq_hd = skb;
+	d->sendq_tl = skb;
 }
 
 static int
@@ -340,13 +377,17 @@ rexmit_timer(ulong vp)
 		if (f->tag != FREETAG && tsince(f->tag) >= timeout) {
 			n = f->waited += timeout;
 			n /= HZ;
-			if (n > MAXWAIT) { /* waited too long.  device failure. */
+			if (n > aoe_deadsecs) { /* waited too long for response */
 				aoedev_downdev(d);
 				break;
 			}
 			rexmit(d, f);
 		}
 	}
+	if (d->flags & DEVFL_KICKME) {
+		d->flags &= ~DEVFL_KICKME;
+		aoecmd_work(d);
+	}
 
 	sl = d->sendq_hd;
 	d->sendq_hd = d->sendq_tl = NULL;
@@ -431,8 +472,8 @@ ataid_complete(struct aoedev *d, unsigned char *id)
 	}
 
 	if (d->ssize != ssize)
-		printk(KERN_INFO "aoe: %012llx e%lu.%lu v%04x has %llu "
-			"sectors\n", (unsigned long long)mac_addr(d->addr),
+		printk(KERN_INFO "aoe: %012llx e%lu.%lu v%04x has %llu sectors\n",
+			(unsigned long long)mac_addr(d->addr),
 			d->aoemajor, d->aoeminor,
 			d->fw_ver, (long long)ssize);
 	d->ssize = ssize;
@@ -442,11 +483,9 @@ ataid_complete(struct aoedev *d, unsigned char *id)
 		d->flags |= DEVFL_NEWSIZE;
 	} else {
 		if (d->flags & DEVFL_GDALLOC) {
-			printk(KERN_INFO "aoe: %s: %s e%lu.%lu, %s\n",
-			       __FUNCTION__,
-			       "can't schedule work for",
+			printk(KERN_ERR "aoe: can't schedule work for e%lu.%lu, %s\n",
 			       d->aoemajor, d->aoeminor,
-			       "it's already on! (This really shouldn't happen).\n");
+			       "it's already on!  This shouldn't happen.\n");
 			return;
 		}
 		d->flags |= DEVFL_GDALLOC;
@@ -460,8 +499,15 @@ calc_rttavg(struct aoedev *d, int rtt)
 	register long n;
 
 	n = rtt;
-	if (n < MINTIMER)
-		n = MINTIMER;
+	if (n < 0) {
+		n = -rtt;
+		if (n < MINTIMER)
+			n = MINTIMER;
+		else if (n > MAXTIMER)
+			n = MAXTIMER;
+		d->mintimer += (n - d->mintimer) >> 1;
+	} else if (n < d->mintimer)
+		n = d->mintimer;
 	else if (n > MAXTIMER)
 		n = MAXTIMER;
 
@@ -474,7 +520,7 @@ void
 aoecmd_ata_rsp(struct sk_buff *skb)
 {
 	struct aoedev *d;
-	struct aoe_hdr *hin;
+	struct aoe_hdr *hin, *hout;
 	struct aoe_atahdr *ahin, *ahout;
 	struct frame *f;
 	struct buf *buf;
@@ -497,8 +543,10 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 
 	spin_lock_irqsave(&d->lock, flags);
 
-	f = getframe(d, be32_to_cpu(hin->tag));
+	n = be32_to_cpu(hin->tag);
+	f = getframe(d, n);
 	if (f == NULL) {
+		calc_rttavg(d, -tsince(n));
 		spin_unlock_irqrestore(&d->lock, flags);
 		snprintf(ebuf, sizeof ebuf,
 			"%15s e%d.%d    tag=%08x@%08lx\n",
@@ -514,26 +562,27 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 	calc_rttavg(d, tsince(f->tag));
 
 	ahin = (struct aoe_atahdr *) (hin+1);
-	ahout = (struct aoe_atahdr *) (f->data + sizeof(struct aoe_hdr));
+	hout = (struct aoe_hdr *) f->skb->mac.raw;
+	ahout = (struct aoe_atahdr *) (hout+1);
 	buf = f->buf;
 
 	if (ahout->cmdstat == WIN_IDENTIFY)
 		d->flags &= ~DEVFL_PAUSE;
 	if (ahin->cmdstat & 0xa9) {	/* these bits cleared on success */
-		printk(KERN_CRIT "aoe: aoecmd_ata_rsp: ata error cmd=%2.2Xh "
-			"stat=%2.2Xh from e%ld.%ld\n", 
+		printk(KERN_ERR
+			"aoe: ata error cmd=%2.2Xh stat=%2.2Xh from e%ld.%ld\n",
 			ahout->cmdstat, ahin->cmdstat,
 			d->aoemajor, d->aoeminor);
 		if (buf)
 			buf->flags |= BUFFL_FAIL;
 	} else {
+		n = ahout->scnt << 9;
 		switch (ahout->cmdstat) {
 		case WIN_READ:
 		case WIN_READ_EXT:
-			n = ahout->scnt << 9;
 			if (skb->len - sizeof *hin - sizeof *ahin < n) {
-				printk(KERN_CRIT "aoe: aoecmd_ata_rsp: runt "
-					"ata data size in read.  skb->len=%d\n",
+				printk(KERN_ERR
+					"aoe: runt data size in read.  skb->len=%d\n",
 					skb->len);
 				/* fail frame f?  just returning will rexmit. */
 				spin_unlock_irqrestore(&d->lock, flags);
@@ -542,22 +591,49 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 			memcpy(f->bufaddr, ahin+1, n);
 		case WIN_WRITE:
 		case WIN_WRITE_EXT:
+			if (f->bcnt -= n) {
+				skb = f->skb;
+				f->bufaddr += n;
+				put_lba(ahout, f->lba += ahout->scnt);
+				n = f->bcnt;
+				if (n > DEFAULTBCNT)
+					n = DEFAULTBCNT;
+				ahout->scnt = n >> 9;
+				if (ahout->aflags & AOEAFL_WRITE) {
+					skb_fill_page_desc(skb, 0,
+						virt_to_page(f->bufaddr),
+						offset_in_page(f->bufaddr), n);
+					skb->len = sizeof *hout + sizeof *ahout + n;
+					skb->data_len = n;
+				}
+				f->tag = newtag(d);
+				hout->tag = cpu_to_be32(f->tag);
+				skb->dev = d->ifp;
+				skb = skb_clone(skb, GFP_ATOMIC);
+				spin_unlock_irqrestore(&d->lock, flags);
+				if (skb)
+					aoenet_xmit(skb);
+				return;
+			}
+			if (n > DEFAULTBCNT)
+				d->lostjumbo = 0;
 			break;
 		case WIN_IDENTIFY:
 			if (skb->len - sizeof *hin - sizeof *ahin < 512) {
-				printk(KERN_INFO "aoe: aoecmd_ata_rsp: runt data size "
-					"in ataid.  skb->len=%d\n", skb->len);
+				printk(KERN_INFO
+					"aoe: runt data size in ataid.  skb->len=%d\n",
+					skb->len);
 				spin_unlock_irqrestore(&d->lock, flags);
 				return;
 			}
 			ataid_complete(d, (char *) (ahin+1));
 			break;
 		default:
-			printk(KERN_INFO "aoe: aoecmd_ata_rsp: unrecognized "
-			       "outbound ata command %2.2Xh for %d.%d\n", 
-			       ahout->cmdstat,
-			       be16_to_cpu(hin->major),
-			       hin->minor);
+			printk(KERN_INFO
+				"aoe: unrecognized ata command %2.2Xh for %d.%d\n",
+				ahout->cmdstat,
+				be16_to_cpu(hin->major),
+				hin->minor);
 		}
 	}
 
@@ -612,33 +688,32 @@ aoecmd_ata_id(struct aoedev *d)
 	struct frame *f;
 	struct sk_buff *skb;
 
-	f = getframe(d, FREETAG);
+	f = freeframe(d);
 	if (f == NULL) {
-		printk(KERN_CRIT "aoe: aoecmd_ata_id: can't get a frame.  "
-			"This shouldn't happen.\n");
+		printk(KERN_ERR "aoe: can't get a frame. This shouldn't happen.\n");
 		return NULL;
 	}
 
 	/* initialize the headers & frame */
-	h = (struct aoe_hdr *) f->data;
+	skb = f->skb;
+	h = (struct aoe_hdr *) skb->mac.raw;
 	ah = (struct aoe_atahdr *) (h+1);
-	f->ndata = sizeof *h + sizeof *ah;
-	memset(h, 0, f->ndata);
+	skb->len = ETH_ZLEN;
+	memset(h, 0, ETH_ZLEN);
 	f->tag = aoehdr_atainit(d, h);
 	f->waited = 0;
-	f->writedatalen = 0;
 
 	/* set up ata header */
 	ah->scnt = 1;
 	ah->cmdstat = WIN_IDENTIFY;
 	ah->lba3 = 0xa0;
 
-	skb = skb_prepare(d, f);
+	skb->dev = d->ifp;
 
 	d->rttavg = MAXTIMER;
 	d->timer.function = rexmit_timer;
 
-	return skb;
+	return skb_clone(skb, GFP_ATOMIC);
 }
  
 void
@@ -648,9 +723,9 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 	struct aoe_hdr *h;
 	struct aoe_cfghdr *ch;
 	ulong flags, sysminor, aoemajor;
-	u16 bufcnt;
 	struct sk_buff *sl;
 	enum { MAXFRAMES = 16 };
+	u16 n;
 
 	h = (struct aoe_hdr *) skb->mac.raw;
 	ch = (struct aoe_cfghdr *) (h+1);
@@ -661,26 +736,25 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 	 */
 	aoemajor = be16_to_cpu(h->major);
 	if (aoemajor == 0xfff) {
-		printk(KERN_CRIT "aoe: aoecmd_cfg_rsp: Warning: shelf "
-			"address is all ones.  Check shelf dip switches\n");
+		printk(KERN_ERR "aoe: Warning: shelf address is all ones.  "
+			"Check shelf dip switches.\n");
 		return;
 	}
 
 	sysminor = SYSMINOR(aoemajor, h->minor);
 	if (sysminor * AOE_PARTITIONS + AOE_PARTITIONS > MINORMASK) {
-		printk(KERN_INFO
-			"aoe: e%ld.%d: minor number too large\n", 
+		printk(KERN_INFO "aoe: e%ld.%d: minor number too large\n",
 			aoemajor, (int) h->minor);
 		return;
 	}
 
-	bufcnt = be16_to_cpu(ch->bufcnt);
-	if (bufcnt > MAXFRAMES)	/* keep it reasonable */
-		bufcnt = MAXFRAMES;
+	n = be16_to_cpu(ch->bufcnt);
+	if (n > MAXFRAMES)	/* keep it reasonable */
+		n = MAXFRAMES;
 
-	d = aoedev_by_sysminor_m(sysminor, bufcnt);
+	d = aoedev_by_sysminor_m(sysminor, n);
 	if (d == NULL) {
-		printk(KERN_INFO "aoe: aoecmd_cfg_rsp: device sysminor_m failure\n");
+		printk(KERN_INFO "aoe: device sysminor_m failure\n");
 		return;
 	}
 
@@ -689,6 +763,20 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 	/* permit device to migrate mac and network interface */
 	d->ifp = skb->dev;
 	memcpy(d->addr, h->src, sizeof d->addr);
+	if (!(d->flags & DEVFL_MAXBCNT)) {
+		n = d->ifp->mtu;
+		n -= sizeof (struct aoe_hdr) + sizeof (struct aoe_atahdr);
+		n /= 512;
+		if (n > ch->scnt)
+			n = ch->scnt;
+		n = n ? n * 512 : DEFAULTBCNT;
+		if (n != d->maxbcnt) {
+			printk(KERN_INFO
+				"aoe: e%ld.%ld: setting %d byte data frames on %s\n",
+				d->aoemajor, d->aoeminor, n, d->ifp->name);
+			d->maxbcnt = n;
+		}
+	}
 
 	/* don't change users' perspective */
 	if (d->nopen && !(d->flags & DEVFL_PAUSE)) {
@@ -696,6 +784,7 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 		return;
 	}
 	d->flags |= DEVFL_PAUSE;	/* force pause */
+	d->mintimer = MINTIMER;
 	d->fw_ver = be16_to_cpu(ch->fwver);
 
 	/* check for already outstanding ataid */
diff --git a/drivers/block/aoe/aoedev.c b/drivers/block/aoe/aoedev.c
index ed4258a62df..6125921bbec 100644
--- a/drivers/block/aoe/aoedev.c
+++ b/drivers/block/aoe/aoedev.c
@@ -1,4 +1,4 @@
-/* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
+/* Copyright (c) 2006 Coraid, Inc.  See COPYING for GPL terms. */
 /*
  * aoedev.c
  * AoE device utility functions; maintains device list.
@@ -20,11 +20,8 @@ aoedev_isbusy(struct aoedev *d)
 	f = d->frames;
 	e = f + d->nframes;
 	do {
-		if (f->tag != FREETAG) {
-			printk(KERN_DEBUG "aoe: %ld.%ld isbusy\n",
-				d->aoemajor, d->aoeminor);
+		if (f->tag != FREETAG)
 			return 1;
-		}
 	} while (++f < e);
 
 	return 0;
@@ -66,22 +63,32 @@ aoedev_newdev(ulong nframes)
 	struct frame *f, *e;
 
 	d = kzalloc(sizeof *d, GFP_ATOMIC);
-	if (d == NULL)
-		return NULL;
 	f = kcalloc(nframes, sizeof *f, GFP_ATOMIC);
-	if (f == NULL) {
-		kfree(d);
+ 	switch (!d || !f) {
+ 	case 0:
+ 		d->nframes = nframes;
+ 		d->frames = f;
+ 		e = f + nframes;
+ 		for (; f<e; f++) {
+ 			f->tag = FREETAG;
+ 			f->skb = new_skb(ETH_ZLEN);
+ 			if (!f->skb)
+ 				break;
+ 		}
+ 		if (f == e)
+ 			break;
+ 		while (f > d->frames) {
+ 			f--;
+ 			dev_kfree_skb(f->skb);
+ 		}
+ 	default:
+ 		if (f)
+ 			kfree(f);
+ 		if (d)
+ 			kfree(d);
 		return NULL;
 	}
-
 	INIT_WORK(&d->work, aoecmd_sleepwork, d);
-
-	d->nframes = nframes;
-	d->frames = f;
-	e = f + nframes;
-	for (; f<e; f++)
-		f->tag = FREETAG;
-
 	spin_lock_init(&d->lock);
 	init_timer(&d->timer);
 	d->timer.data = (ulong) d;
@@ -114,6 +121,7 @@ aoedev_downdev(struct aoedev *d)
 			mempool_free(buf, d->bufpool);
 			bio_endio(bio, bio->bi_size, -EIO);
 		}
+		skb_shinfo(f->skb)->nr_frags = f->skb->data_len = 0;
 	}
 	d->inprocess = NULL;
 
@@ -148,7 +156,7 @@ aoedev_by_sysminor_m(ulong sysminor, ulong bufcnt)
 		d = aoedev_newdev(bufcnt);
 	 	if (d == NULL) {
 			spin_unlock_irqrestore(&devlist_lock, flags);
-			printk(KERN_INFO "aoe: aoedev_set: aoedev_newdev failure.\n");
+			printk(KERN_INFO "aoe: aoedev_newdev failure.\n");
 			return NULL;
 		}
 		d->sysminor = sysminor;
@@ -163,11 +171,19 @@ aoedev_by_sysminor_m(ulong sysminor, ulong bufcnt)
 static void
 aoedev_freedev(struct aoedev *d)
 {
+	struct frame *f, *e;
+
 	if (d->gd) {
 		aoedisk_rm_sysfs(d);
 		del_gendisk(d->gd);
 		put_disk(d->gd);
 	}
+	f = d->frames;
+	e = f + d->nframes;
+	for (; f<e; f++) {
+		skb_shinfo(f->skb)->nr_frags = 0;
+		dev_kfree_skb(f->skb);
+	}
 	kfree(d->frames);
 	if (d->bufpool)
 		mempool_destroy(d->bufpool);
diff --git a/drivers/block/aoe/aoemain.c b/drivers/block/aoe/aoemain.c
index de08491ebe6..a04b7d61329 100644
--- a/drivers/block/aoe/aoemain.c
+++ b/drivers/block/aoe/aoemain.c
@@ -1,4 +1,4 @@
-/* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
+/* Copyright (c) 2006 Coraid, Inc.  See COPYING for GPL terms. */
 /*
  * aoemain.c
  * Module initialization routines, discover timer
@@ -84,13 +84,11 @@ aoe_init(void)
 		goto net_fail;
 	ret = register_blkdev(AOE_MAJOR, DEVICE_NAME);
 	if (ret < 0) {
-		printk(KERN_ERR "aoe: aoeblk_init: can't register major\n");
+		printk(KERN_ERR "aoe: can't register major\n");
 		goto blkreg_fail;
 	}
 
-	printk(KERN_INFO
-	       "aoe: aoe_init: AoE v%s initialised.\n",
-	       VERSION);
+	printk(KERN_INFO "aoe: AoE v%s initialised.\n", VERSION);
 	discover_timer(TINIT);
 	return 0;
 
@@ -103,7 +101,7 @@ aoe_init(void)
  chr_fail:
 	aoedev_exit();
 	
-	printk(KERN_INFO "aoe: aoe_init: initialisation failure.\n");
+	printk(KERN_INFO "aoe: initialisation failure.\n");
 	return ret;
 }
 
diff --git a/drivers/block/aoe/aoenet.c b/drivers/block/aoe/aoenet.c
index c1434ed1188..9626e0f5da9 100644
--- a/drivers/block/aoe/aoenet.c
+++ b/drivers/block/aoe/aoenet.c
@@ -1,4 +1,4 @@
-/* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
+/* Copyright (c) 2006 Coraid, Inc.  See COPYING for GPL terms. */
 /*
  * aoenet.c
  * Ethernet portion of AoE driver
@@ -74,7 +74,7 @@ set_aoe_iflist(const char __user *user_str, size_t size)
 		return -EINVAL;
 
 	if (copy_from_user(aoe_iflist, user_str, size)) {
-		printk(KERN_INFO "aoe: %s: copy from user failed\n", __FUNCTION__);
+		printk(KERN_INFO "aoe: copy from user failed\n");
 		return -EFAULT;
 	}
 	aoe_iflist[size] = 0x00;
@@ -132,8 +132,7 @@ aoenet_rcv(struct sk_buff *skb, struct net_device *ifp, struct packet_type *pt,
 		if (n > NECODES)
 			n = 0;
 		if (net_ratelimit())
-			printk(KERN_ERR "aoe: aoenet_rcv: error packet from %d.%d; "
-			       "ecode=%d '%s'\n",
+			printk(KERN_ERR "aoe: error packet from %d.%d; ecode=%d '%s'\n",
 			       be16_to_cpu(h->major), h->minor, 
 			       h->err, aoe_errlist[n]);
 		goto exit;
@@ -147,7 +146,7 @@ aoenet_rcv(struct sk_buff *skb, struct net_device *ifp, struct packet_type *pt,
 		aoecmd_cfg_rsp(skb);
 		break;
 	default:
-		printk(KERN_INFO "aoe: aoenet_rcv: unknown cmd %d\n", h->cmd);
+		printk(KERN_INFO "aoe: unknown cmd %d\n", h->cmd);
 	}
 exit:
 	dev_kfree_skb(skb);