Re: [PATCH 2.6.11] aoe [1/12]: remove too-low cap on minor number

Re: [PATCH 2.6.11] aoe [1/12]: remove too-low cap on minor number

[Greg KH]

I've applied 11 of these 12 patches (the one from Randy was already
included) to my trees.

thanks,

greg k-h

Re: [PATCH 2.6.11] aoe [1/12]: remove too-low cap on minor number

[Ed L Cashin]

Greg KH <greg@kroah.com> writes:

> I've applied 11 of these 12 patches (the one from Randy was already
> included) to my trees.

Those haven't gone to the lkml yet.  I'll post them through gmane.
(The original postings never made it to the list, and I can't get in
touch with the lkml admins, but gmane should work.)

-- 
  Ed L Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [4/12]: handle distros that have a udev rules file instead of dir

[Ed L. Cashin]

handle distros that have a udev rules file instead of dir

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/Documentation/aoe/udev-install.sh b/Documentation/aoe/udev-install.sh
--- a/Documentation/aoe/udev-install.sh	2005-03-10 11:59:55.000000000 -0500
+++ b/Documentation/aoe/udev-install.sh	2005-03-10 12:19:18.000000000 -0500
@@ -23,4 +23,8 @@ fi
 # /etc/udev/rules.d
 #
 rules_d="`sed -n '/^udev_rules=/{ s!udev_rules=!!; s!\"!!g; p; }' $conf`"
-test "$rules_d" && sh -xc "cp `dirname $0`/udev.txt $rules_d/60-aoe.rules"
+if test -z "$rules_d" || test ! -d "$rules_d"; then
+	echo "$me Error: cannot find udev rules directory" 1>&2
+	exit 1
+fi
+sh -xc "cp `dirname $0`/udev.txt $rules_d/60-aoe.rules"


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [1/12]: remove too-low cap on minor number

[ecashin]

remove too-low cap on minor number

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
--- a/drivers/block/aoe/aoe.h	2005-03-10 11:59:55.000000000 -0500
+++ b/drivers/block/aoe/aoe.h	2005-03-10 12:19:04.000000000 -0500
@@ -2,9 +2,14 @@
 #define VERSION "5"
 #define AOE_MAJOR 152
 #define DEVICE_NAME "aoe"
+
+/* set AOE_PARTITIONS to 1 to use whole-disks only
+ * default is 16, which is 15 partitions plus the whole disk
+ */
 #ifndef AOE_PARTITIONS
 #define AOE_PARTITIONS 16
 #endif
+
 #define SYSMINOR(aoemajor, aoeminor) ((aoemajor) * 10 + (aoeminor))
 #define AOEMAJOR(sysminor) ((sysminor) / 10)
 #define AOEMINOR(sysminor) ((sysminor) % 10)
diff -uprN a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
--- a/drivers/block/aoe/aoecmd.c	2005-03-10 11:59:55.000000000 -0500
+++ b/drivers/block/aoe/aoecmd.c	2005-03-10 12:19:04.000000000 -0500
@@ -577,7 +577,7 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 	struct aoe_cfghdr *ch;
 	ulong flags, bufcnt, sysminor, aoemajor;
 	struct sk_buff *sl;
-	enum { MAXFRAMES = 8, MAXSYSMINOR = 255 };
+	enum { MAXFRAMES = 8 };
 
 	h = (struct aoe_hdr *) skb->mac.raw;
 	ch = (struct aoe_cfghdr *) (h+1);
@@ -594,9 +594,10 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 	}
 
 	sysminor = SYSMINOR(aoemajor, h->minor);
-	if (sysminor > MAXSYSMINOR) {
-		printk(KERN_INFO "aoe: aoecmd_cfg_rsp: sysminor %ld too "
-			"large\n", sysminor);
+	if (sysminor * AOE_PARTITIONS + AOE_PARTITIONS > MINORMASK) {
+		printk(KERN_INFO
+			"aoe: e%ld.%d: minor number too large\n", 
+			aoemajor, (int) h->minor);
 		return;
 	}
 


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [2/12]: allow multiple aoe devices with same MAC addr

[ecashin]

allow multiple aoe devices with same MAC addr

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
--- a/drivers/block/aoe/aoe.h	2005-03-10 12:19:04.000000000 -0500
+++ b/drivers/block/aoe/aoe.h	2005-03-10 12:19:11.000000000 -0500
@@ -156,7 +156,7 @@ void aoecmd_cfg_rsp(struct sk_buff *);
 
 int aoedev_init(void);
 void aoedev_exit(void);
-struct aoedev *aoedev_bymac(unsigned char *);
+struct aoedev *aoedev_by_aoeaddr(int maj, int min);
 void aoedev_downdev(struct aoedev *d);
 struct aoedev *aoedev_set(ulong, unsigned char *, struct net_device *, ulong);
 int aoedev_busy(void);
diff -uprN a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
--- a/drivers/block/aoe/aoecmd.c	2005-03-10 12:19:04.000000000 -0500
+++ b/drivers/block/aoe/aoecmd.c	2005-03-10 12:19:11.000000000 -0500
@@ -380,14 +380,15 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 	register long n;
 	ulong flags;
 	char ebuf[128];
-	
+	u16 aoemajor;
+
 	hin = (struct aoe_hdr *) skb->mac.raw;
-	d = aoedev_bymac(hin->src);
+	aoemajor = __be16_to_cpu(*((u16 *) hin->major));
+	d = aoedev_by_aoeaddr(aoemajor, hin->minor);
 	if (d == NULL) {
 		snprintf(ebuf, sizeof ebuf, "aoecmd_ata_rsp: ata response "
 			"for unknown device %d.%d\n",
-			 __be16_to_cpu(*((u16 *) hin->major)),
-			hin->minor);
+			 aoemajor, hin->minor);
 		aoechr_error(ebuf);
 		return;
 	}
diff -uprN a/drivers/block/aoe/aoedev.c b/drivers/block/aoe/aoedev.c
--- a/drivers/block/aoe/aoedev.c	2005-03-07 17:37:14.000000000 -0500
+++ b/drivers/block/aoe/aoedev.c	2005-03-10 12:19:11.000000000 -0500
@@ -13,7 +13,7 @@ static struct aoedev *devlist;
 static spinlock_t devlist_lock;
 
 struct aoedev *
-aoedev_bymac(unsigned char *macaddr)
+aoedev_by_aoeaddr(int maj, int min)
 {
 	struct aoedev *d;
 	ulong flags;
@@ -21,7 +21,7 @@ aoedev_bymac(unsigned char *macaddr)
 	spin_lock_irqsave(&devlist_lock, flags);
 
 	for (d=devlist; d; d=d->next)
-		if (!memcmp(d->addr, macaddr, 6))
+		if (d->aoemajor == maj && d->aoeminor == min)
 			break;
 
 	spin_unlock_irqrestore(&devlist_lock, flags);
@@ -125,7 +125,6 @@ aoedev_set(ulong sysminor, unsigned char
 	d->ifp = ifp;
 
 	if (d->sysminor != sysminor
-	|| memcmp(d->addr, addr, sizeof d->addr)
 	|| (d->flags & DEVFL_UP) == 0) {
 		aoedev_downdev(d); /* flushes outstanding frames */
 		memcpy(d->addr, addr, sizeof d->addr);


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [3/12]: update driver version to 6

[ecashin]

update driver version to 6

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
--- a/drivers/block/aoe/aoe.h	2005-03-10 12:19:11.000000000 -0500
+++ b/drivers/block/aoe/aoe.h	2005-03-10 12:19:14.000000000 -0500
@@ -1,5 +1,5 @@
 /* Copyright (c) 2004 Coraid, Inc.  See COPYING for GPL terms. */
-#define VERSION "5"
+#define VERSION "6"
 #define AOE_MAJOR 152
 #define DEVICE_NAME "aoe"
 


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [4/12]: handle distros that have a udev rules file instead of dir

[ecashin]

handle distros that have a udev rules file instead of dir

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/Documentation/aoe/udev-install.sh b/Documentation/aoe/udev-install.sh
--- a/Documentation/aoe/udev-install.sh	2005-03-10 11:59:55.000000000 -0500
+++ b/Documentation/aoe/udev-install.sh	2005-03-10 12:19:18.000000000 -0500
@@ -23,4 +23,8 @@ fi
 # /etc/udev/rules.d
 #
 rules_d="`sed -n '/^udev_rules=/{ s!udev_rules=!!; s!\"!!g; p; }' $conf`"
-test "$rules_d" && sh -xc "cp `dirname $0`/udev.txt $rules_d/60-aoe.rules"
+if test -z "$rules_d" || test ! -d "$rules_d"; then
+	echo "$me Error: cannot find udev rules directory" 1>&2
+	exit 1
+fi
+sh -xc "cp `dirname $0`/udev.txt $rules_d/60-aoe.rules"


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [5/12]: don't try to free null bufpool

[ecashin]

don't try to free null bufpool

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoedev.c b/drivers/block/aoe/aoedev.c
--- a/drivers/block/aoe/aoedev.c	2005-03-10 12:19:11.000000000 -0500
+++ b/drivers/block/aoe/aoedev.c	2005-03-10 12:19:25.000000000 -0500
@@ -146,7 +146,8 @@ aoedev_freedev(struct aoedev *d)
 		put_disk(d->gd);
 	}
 	kfree(d->frames);
-	mempool_destroy(d->bufpool);
+	if (d->bufpool)
+		mempool_destroy(d->bufpool);
 	kfree(d);
 }
 


-- 
  Ed L. Cashin <ecashin@coraid.com>

Re: [PATCH 2.6.11] aoe [5/12]: don't try to free null bufpool

[Arjan van de Ven]

On Thu, 2005-03-24 at 07:17 -0800, ecashin@noserose.net wrote:
> don't try to free null bufpool

in linux there is a "rule" that all memory free routines are supposed to
also accept NULL as argument, so I think this patch is not needed (and
even wrong)

Re: [PATCH 2.6.11] aoe [5/12]: don't try to free null bufpool

[ecashin]

Arjan van de Ven <arjan@infradead.org> writes:

> On Thu, 2005-03-24 at 07:17 -0800, ecashin@noserose.net wrote:
>> don't try to free null bufpool
>
> in linux there is a "rule" that all memory free routines are supposed to
> also accept NULL as argument, so I think this patch is not needed (and
> even wrong)
>

Hmm.  The mm/mempool.c:mempool_destroy function immediately
dereferences the pointer passed to it:

void mempool_destroy(mempool_t *pool)
{
	if (pool->curr_nr != pool->min_nr)
		BUG();		/* There were outstanding elements */
	free_pool(pool);
}

... so I'm not sure mempool_destroy fits the rule.  Are you suggesting
that the patch should instead modify mempool_destroy?

-- 
  Ed L Cashin <ecashin@coraid.com>

Re: [PATCH 2.6.11] aoe [5/12]: don't try to free null bufpool

[Arjan van de Ven]

On Thu, 2005-03-24 at 09:04 -0800, ecashin@noserose.net wrote:
> Arjan van de Ven <arjan@infradead.org> writes:
> 
> > On Thu, 2005-03-24 at 07:17 -0800, ecashin@noserose.net wrote:
> >> don't try to free null bufpool
> >
> > in linux there is a "rule" that all memory free routines are supposed to
> > also accept NULL as argument, so I think this patch is not needed (and
> > even wrong)
> >
> 
> Hmm.  The mm/mempool.c:mempool_destroy function immediately
> dereferences the pointer passed to it:
> 
> void mempool_destroy(mempool_t *pool)
> {
> 	if (pool->curr_nr != pool->min_nr)
> 		BUG();		/* There were outstanding elements */
> 	free_pool(pool);
> }
> 
> ... so I'm not sure mempool_destroy fits the rule.  Are you suggesting
> that the patch should instead modify mempool_destroy?

hmm perhaps... Jens?

Re: [PATCH 2.6.11] aoe [5/12]: don't try to free null bufpool

[Jesper Juhl]

On Thu, 24 Mar 2005, Arjan van de Ven wrote:

> On Thu, 2005-03-24 at 09:04 -0800, ecashin@noserose.net wrote:
> > Arjan van de Ven <arjan@infradead.org> writes:
> > 
> > > On Thu, 2005-03-24 at 07:17 -0800, ecashin@noserose.net wrote:
> > >> don't try to free null bufpool
> > >
> > > in linux there is a "rule" that all memory free routines are supposed to
> > > also accept NULL as argument, so I think this patch is not needed (and
> > > even wrong)
> > >
> > 
> > Hmm.  The mm/mempool.c:mempool_destroy function immediately
> > dereferences the pointer passed to it:
> > 
> > void mempool_destroy(mempool_t *pool)
> > {
> > 	if (pool->curr_nr != pool->min_nr)
> > 		BUG();		/* There were outstanding elements */
> > 	free_pool(pool);
> > }
> > 
> > ... so I'm not sure mempool_destroy fits the rule.  Are you suggesting
> > that the patch should instead modify mempool_destroy?
> 
> hmm perhaps... Jens?
> 

Having mempool_destroy() be the one that checks seems safer, then callers 
won't forget to check - easier to just check in one place.
If that's what you want, then here's a patch. If this is acceptable I can 
create another one that removes the (then pointless) NULL checks from all 
callers - let me know if that's wanted.

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>

--- linux-2.6.12-rc1-mm3-orig/mm/mempool.c	2005-03-21 23:12:43.000000000 +0100
+++ linux-2.6.12-rc1-mm3/mm/mempool.c	2005-03-25 15:34:04.000000000 +0100
@@ -176,6 +176,8 @@ EXPORT_SYMBOL(mempool_resize);
  */
 void mempool_destroy(mempool_t *pool)
 {
+	if (!pool)
+		return;
 	if (pool->curr_nr != pool->min_nr)
 		BUG();		/* There were outstanding elements */
 	free_pool(pool);

Re: [PATCH 2.6.11] aoe [5/12]: don't try to free null bufpool

[Jens Axboe]

On Thu, Mar 24 2005, Arjan van de Ven wrote:
> On Thu, 2005-03-24 at 09:04 -0800, ecashin@noserose.net wrote:
> > Arjan van de Ven <arjan@infradead.org> writes:
> > 
> > > On Thu, 2005-03-24 at 07:17 -0800, ecashin@noserose.net wrote:
> > >> don't try to free null bufpool
> > >
> > > in linux there is a "rule" that all memory free routines are supposed to
> > > also accept NULL as argument, so I think this patch is not needed (and
> > > even wrong)
> > >
> > 
> > Hmm.  The mm/mempool.c:mempool_destroy function immediately
> > dereferences the pointer passed to it:
> > 
> > void mempool_destroy(mempool_t *pool)
> > {
> > 	if (pool->curr_nr != pool->min_nr)
> > 		BUG();		/* There were outstanding elements */
> > 	free_pool(pool);
> > }
> > 
> > ... so I'm not sure mempool_destroy fits the rule.  Are you suggesting
> > that the patch should instead modify mempool_destroy?
> 
> hmm perhaps... Jens?

Not really my call, but I agree we should make mempool_destroy()
resilient against !pool to follow the path of least surprise.

-- 
Jens Axboe

[PATCH 2.6.11] aoe [6/12]: Alexey Dobriyan sparse cleanup

[ecashin]

Alexey Dobriyan sparse cleanup

Signed-off-by: Alexey Dobriyan <adobriyan@mail.ru>
Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
--- a/drivers/block/aoe/aoe.h	2005-03-10 12:19:14.000000000 -0500
+++ b/drivers/block/aoe/aoe.h	2005-03-10 12:19:27.000000000 -0500
@@ -39,13 +39,13 @@ enum {
 struct aoe_hdr {
 	unsigned char dst[6];
 	unsigned char src[6];
-	unsigned char type[2];
+	__be16 type;
 	unsigned char verfl;
 	unsigned char err;
-	unsigned char major[2];
+	__be16 major;
 	unsigned char minor;
 	unsigned char cmd;
-	unsigned char tag[4];
+	__be32 tag;
 };
 
 struct aoe_atahdr {
@@ -63,8 +63,8 @@ struct aoe_atahdr {
 };
 
 struct aoe_cfghdr {
-	unsigned char bufcnt[2];
-	unsigned char fwver[2];
+	__be16 bufcnt;
+	__be16 fwver;
 	unsigned char res;
 	unsigned char aoeccmd;
 	unsigned char cslen[2];
diff -uprN a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
--- a/drivers/block/aoe/aoecmd.c	2005-03-10 12:19:11.000000000 -0500
+++ b/drivers/block/aoe/aoecmd.c	2005-03-10 12:19:27.000000000 -0500
@@ -90,19 +90,16 @@ newtag(struct aoedev *d)
 static int
 aoehdr_atainit(struct aoedev *d, struct aoe_hdr *h)
 {
-	u16 type = __constant_cpu_to_be16(ETH_P_AOE);
-	u16 aoemajor = __cpu_to_be16(d->aoemajor);
 	u32 host_tag = newtag(d);
-	u32 tag = __cpu_to_be32(host_tag);
 
 	memcpy(h->src, d->ifp->dev_addr, sizeof h->src);
 	memcpy(h->dst, d->addr, sizeof h->dst);
-	memcpy(h->type, &type, sizeof type);
+	h->type = __constant_cpu_to_be16(ETH_P_AOE);
 	h->verfl = AOE_HVER;
-	memcpy(h->major, &aoemajor, sizeof aoemajor);
+	h->major = cpu_to_be16(d->aoemajor);
 	h->minor = d->aoeminor;
 	h->cmd = AOECMD_ATA;
-	memcpy(h->tag, &tag, sizeof tag);
+	h->tag = cpu_to_be32(host_tag);
 
 	return host_tag;
 }
@@ -215,7 +212,6 @@ rexmit(struct aoedev *d, struct frame *f
 	struct aoe_hdr *h;
 	char buf[128];
 	u32 n;
-	u32 net_tag;
 
 	n = newtag(d);
 
@@ -227,8 +223,7 @@ rexmit(struct aoedev *d, struct frame *f
 
 	h = (struct aoe_hdr *) f->data;
 	f->tag = n;
-	net_tag = __cpu_to_be32(n);
-	memcpy(h->tag, &net_tag, sizeof net_tag);
+	h->tag = cpu_to_be32(n);
 
 	skb = skb_prepare(d, f);
 	if (skb) {
@@ -308,16 +303,16 @@ ataid_complete(struct aoedev *d, unsigne
 	u16 n;
 
 	/* word 83: command set supported */
-	n = __le16_to_cpu(*((u16 *) &id[83<<1]));
+	n = le16_to_cpup((__le16 *) &id[83<<1]);
 
 	/* word 86: command set/feature enabled */
-	n |= __le16_to_cpu(*((u16 *) &id[86<<1]));
+	n |= le16_to_cpup((__le16 *) &id[86<<1]);
 
 	if (n & (1<<10)) {	/* bit 10: LBA 48 */
 		d->flags |= DEVFL_EXT;
 
 		/* word 100: number lba48 sectors */
-		ssize = __le64_to_cpu(*((u64 *) &id[100<<1]));
+		ssize = le64_to_cpup((__le64 *) &id[100<<1]);
 
 		/* set as in ide-disk.c:init_idedisk_capacity */
 		d->geo.cylinders = ssize;
@@ -328,12 +323,12 @@ ataid_complete(struct aoedev *d, unsigne
 		d->flags &= ~DEVFL_EXT;
 
 		/* number lba28 sectors */
-		ssize = __le32_to_cpu(*((u32 *) &id[60<<1]));
+		ssize = le32_to_cpup((__le32 *) &id[60<<1]);
 
 		/* NOTE: obsolete in ATA 6 */
-		d->geo.cylinders = __le16_to_cpu(*((u16 *) &id[54<<1]));
-		d->geo.heads = __le16_to_cpu(*((u16 *) &id[55<<1]));
-		d->geo.sectors = __le16_to_cpu(*((u16 *) &id[56<<1]));
+		d->geo.cylinders = le16_to_cpup((__le16 *) &id[54<<1]);
+		d->geo.heads = le16_to_cpup((__le16 *) &id[55<<1]);
+		d->geo.sectors = le16_to_cpup((__le16 *) &id[56<<1]);
 	}
 	d->ssize = ssize;
 	d->geo.start = 0;
@@ -383,7 +378,7 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 	u16 aoemajor;
 
 	hin = (struct aoe_hdr *) skb->mac.raw;
-	aoemajor = __be16_to_cpu(*((u16 *) hin->major));
+	aoemajor = be16_to_cpu(hin->major);
 	d = aoedev_by_aoeaddr(aoemajor, hin->minor);
 	if (d == NULL) {
 		snprintf(ebuf, sizeof ebuf, "aoecmd_ata_rsp: ata response "
@@ -395,15 +390,15 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 
 	spin_lock_irqsave(&d->lock, flags);
 
-	f = getframe(d, __be32_to_cpu(*((u32 *) hin->tag)));
+	f = getframe(d, be32_to_cpu(hin->tag));
 	if (f == NULL) {
 		spin_unlock_irqrestore(&d->lock, flags);
 		snprintf(ebuf, sizeof ebuf,
 			"%15s e%d.%d    tag=%08x@%08lx\n",
 			"unexpected rsp",
-			__be16_to_cpu(*((u16 *) hin->major)),
+			be16_to_cpu(hin->major),
 			hin->minor,
-			__be32_to_cpu(*((u32 *) hin->tag)),
+			be32_to_cpu(hin->tag),
 			jiffies);
 		aoechr_error(ebuf);
 		return;
@@ -453,7 +448,7 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 			printk(KERN_INFO "aoe: aoecmd_ata_rsp: unrecognized "
 			       "outbound ata command %2.2Xh for %d.%d\n", 
 			       ahout->cmdstat,
-			       __be16_to_cpu(*((u16 *) hin->major)),
+			       be16_to_cpu(hin->major),
 			       hin->minor);
 		}
 	}
@@ -487,8 +482,6 @@ aoecmd_cfg(ushort aoemajor, unsigned cha
 	struct aoe_cfghdr *ch;
 	struct sk_buff *skb, *sl;
 	struct net_device *ifp;
-	u16 aoe_type = __constant_cpu_to_be16(ETH_P_AOE);
-	u16 net_aoemajor = __cpu_to_be16(aoemajor);
 
 	sl = NULL;
 
@@ -508,9 +501,9 @@ aoecmd_cfg(ushort aoemajor, unsigned cha
 
 		memset(h->dst, 0xff, sizeof h->dst);
 		memcpy(h->src, ifp->dev_addr, sizeof h->src);
-		memcpy(h->type, &aoe_type, sizeof aoe_type);
+		h->type = __constant_cpu_to_be16(ETH_P_AOE);
 		h->verfl = AOE_HVER;
-		memcpy(h->major, &net_aoemajor, sizeof net_aoemajor);
+		h->major = cpu_to_be16(aoemajor);
 		h->minor = aoeminor;
 		h->cmd = AOECMD_CFG;
 
@@ -576,7 +569,8 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 	struct aoedev *d;
 	struct aoe_hdr *h;
 	struct aoe_cfghdr *ch;
-	ulong flags, bufcnt, sysminor, aoemajor;
+	ulong flags, sysminor, aoemajor;
+	u16 bufcnt;
 	struct sk_buff *sl;
 	enum { MAXFRAMES = 8 };
 
@@ -587,7 +581,7 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 	 * Enough people have their dip switches set backwards to
 	 * warrant a loud message for this special case.
 	 */
-	aoemajor = __be16_to_cpu(*((u16 *) h->major));
+	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");
@@ -602,7 +596,7 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 		return;
 	}
 
-	bufcnt = __be16_to_cpu(*((u16 *) ch->bufcnt));
+	bufcnt = be16_to_cpu(ch->bufcnt);
 	if (bufcnt > MAXFRAMES)	/* keep it reasonable */
 		bufcnt = MAXFRAMES;
 
@@ -619,7 +613,7 @@ aoecmd_cfg_rsp(struct sk_buff *skb)
 		return;
 	}
 
-	d->fw_ver = __be16_to_cpu(*((u16 *) ch->fwver));
+	d->fw_ver = be16_to_cpu(ch->fwver);
 
 	/* we get here only if the device is new */
 	sl = aoecmd_ata_id(d);
diff -uprN a/drivers/block/aoe/aoenet.c b/drivers/block/aoe/aoenet.c
--- a/drivers/block/aoe/aoenet.c	2005-03-07 17:37:14.000000000 -0500
+++ b/drivers/block/aoe/aoenet.c	2005-03-10 12:19:27.000000000 -0500
@@ -69,7 +69,7 @@ set_aoe_iflist(const char __user *user_s
 u64
 mac_addr(char addr[6])
 {
-	u64 n = 0;
+	__be64 n = 0;
 	char *p = (char *) &n;
 
 	memcpy(p + 2, addr, 6);	/* (sizeof addr != 6) */
@@ -108,7 +108,7 @@ static int
 aoenet_rcv(struct sk_buff *skb, struct net_device *ifp, struct packet_type *pt)
 {
 	struct aoe_hdr *h;
-	ulong n;
+	u32 n;
 
 	skb = skb_check(skb);
 	if (!skb)
@@ -121,7 +121,7 @@ aoenet_rcv(struct sk_buff *skb, struct n
 	skb_push(skb, ETH_HLEN);	/* (1) */
 
 	h = (struct aoe_hdr *) skb->mac.raw;
-	n = __be32_to_cpu(*((u32 *) h->tag));
+	n = be32_to_cpu(h->tag);
 	if ((h->verfl & AOEFL_RSP) == 0 || (n & 1<<31))
 		goto exit;
 
@@ -132,7 +132,7 @@ aoenet_rcv(struct sk_buff *skb, struct n
 		if (net_ratelimit())
 			printk(KERN_ERR "aoe: aoenet_rcv: error packet from %d.%d; "
 			       "ecode=%d '%s'\n",
-			       __be16_to_cpu(*((u16 *) h->major)), h->minor, 
+			       be16_to_cpu(h->major), h->minor, 
 			       h->err, aoe_errlist[n]);
 		goto exit;
 	}


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[ecashin]

support configuration of AOE_PARTITIONS from Kconfig

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/Kconfig b/drivers/block/Kconfig
--- a/drivers/block/Kconfig	2005-03-07 17:37:58.000000000 -0500
+++ b/drivers/block/Kconfig	2005-03-10 12:19:54.000000000 -0500
@@ -506,4 +506,19 @@ config ATA_OVER_ETH
 	This driver provides Support for ATA over Ethernet block
 	devices like the Coraid EtherDrive (R) Storage Blade.
 
+config AOE_PARTITIONS
+	int "Partitions per AoE device" if ATA_OVER_ETH
+	default "16"
+	help
+	  The default is to support 16 partitions per aoe device. Some
+	  systems lack good support for devices with large minor
+	  numbers.
+
+	  Such systems will be able to use more aoe disks when
+	  AOE_PARTITIONS is set to one, but you won't be able to
+	  partition the disks, and you must make sure your device
+	  nodes are created to work with the value you select.
+
+	  If unsure, use 16.
+
 endmenu
diff -uprN a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
--- a/drivers/block/aoe/aoe.h	2005-03-10 12:19:27.000000000 -0500
+++ b/drivers/block/aoe/aoe.h	2005-03-10 12:19:54.000000000 -0500
@@ -6,8 +6,9 @@
 /* set AOE_PARTITIONS to 1 to use whole-disks only
  * default is 16, which is 15 partitions plus the whole disk
  */
-#ifndef AOE_PARTITIONS
-#define AOE_PARTITIONS 16
+#define AOE_PARTITIONS CONFIG_AOE_PARTITIONS
+#if AOE_PARTITIONS < 1
+#error AOE_PARTITIONS less than one
 #endif
 
 #define SYSMINOR(aoemajor, aoeminor) ((aoemajor) * 10 + (aoeminor))


-- 
  Ed L. Cashin <ecashin@coraid.com>

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Christoph Hellwig]

On Thu, Mar 24, 2005 at 07:21:28AM -0800, ecashin@noserose.net wrote:
> 
> support configuration of AOE_PARTITIONS from Kconfig
> 
> Signed-off-by: Ed L. Cashin <ecashin@coraid.com>
> 
> diff -uprN a/drivers/block/Kconfig b/drivers/block/Kconfig
> --- a/drivers/block/Kconfig	2005-03-07 17:37:58.000000000 -0500
> +++ b/drivers/block/Kconfig	2005-03-10 12:19:54.000000000 -0500
> @@ -506,4 +506,19 @@ config ATA_OVER_ETH
>  	This driver provides Support for ATA over Ethernet block
>  	devices like the Coraid EtherDrive (R) Storage Blade.
>  
> +config AOE_PARTITIONS
> +	int "Partitions per AoE device" if ATA_OVER_ETH
> +	default "16"
> +	help
> +	  The default is to support 16 partitions per aoe device. Some
> +	  systems lack good support for devices with large minor
> +	  numbers.
> +
> +	  Such systems will be able to use more aoe disks when
> +	  AOE_PARTITIONS is set to one, but you won't be able to
> +	  partition the disks, and you must make sure your device
> +	  nodes are created to work with the value you select.
> +
> +	  If unsure, use 16.
> +

NACK.  this changes devices nodes based on a compile-time option.  Just
tell people to update their userland to a 2.6-copatible version.

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Ed L Cashin]

Christoph Hellwig <hch@infradead.org> writes:

> On Thu, Mar 24, 2005 at 07:21:28AM -0800, ecashin@noserose.net wrote:
>> 
>> support configuration of AOE_PARTITIONS from Kconfig
>> 
>> Signed-off-by: Ed L. Cashin <ecashin@coraid.com>
>> 
>> diff -uprN a/drivers/block/Kconfig b/drivers/block/Kconfig
>> --- a/drivers/block/Kconfig	2005-03-07 17:37:58.000000000 -0500
>> +++ b/drivers/block/Kconfig	2005-03-10 12:19:54.000000000 -0500
>> @@ -506,4 +506,19 @@ config ATA_OVER_ETH
>>  	This driver provides Support for ATA over Ethernet block
>>  	devices like the Coraid EtherDrive (R) Storage Blade.
>>  
>> +config AOE_PARTITIONS
>> +	int "Partitions per AoE device" if ATA_OVER_ETH
>> +	default "16"
>> +	help
>> +	  The default is to support 16 partitions per aoe device. Some
>> +	  systems lack good support for devices with large minor
>> +	  numbers.
>> +
>> +	  Such systems will be able to use more aoe disks when
>> +	  AOE_PARTITIONS is set to one, but you won't be able to
>> +	  partition the disks, and you must make sure your device
>> +	  nodes are created to work with the value you select.
>> +
>> +	  If unsure, use 16.
>> +
>
> NACK.  this changes devices nodes based on a compile-time option.  

I'm not sure I follow.  This configuration option sets the number of
partitions per device in the driver.  It doesn't create device nodes.

If the user has udev, then the device nodes are created correctly (on
Fedora Core 3), so that if the driver is configured with 1 partition
per device, the minor numbers for the disks are low.  

The folks I've talked to who aren't using udev but want one partition
per device already know that they have to re-create their device
files.

> Just tell people to update their userland to a 2.6-copatible
> version.

Even if the glibc, coreutils, etc., get it right, some programs try to
parse the device node bits themselves and fail to find all the minor
number bits.

Making this configurable makes it possible for a debian sarge user or
a Slackware 10 user to run a 2.6.11 kernel and use up to 256 disks.
Even a Fedora Core 3 user has an mdadm that balks at minor numbers
like 1120.  By using one partition per device, I can use FC3's mdadm
and have everything work.

The AoE users started doing this themselves, but this configuration
option allows non-C-programmers to do the same.  It's helpful during
the transitional period and should be removed when the userland
software that people are actually running has caught up.

-- 
  Ed L Cashin <ecashin@coraid.com>

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Christoph Hellwig]

On Tue, Mar 29, 2005 at 11:06:16AM -0500, Ed L Cashin wrote:
> >
> > NACK.  this changes devices nodes based on a compile-time option.  
> 
> I'm not sure I follow.  This configuration option sets the number of
> partitions per device in the driver.  It doesn't create device nodes.
> 
> If the user has udev, then the device nodes are created correctly (on
> Fedora Core 3), so that if the driver is configured with 1 partition
> per device, the minor numbers for the disks are low.  
> 
> The folks I've talked to who aren't using udev but want one partition
> per device already know that they have to re-create their device
> files.

It changes a kernel ABI, so people that have different config options
set can't use the same userland.  It's a really big no-go.

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Ed L Cashin]

Christoph Hellwig <hch@infradead.org> writes:

> On Tue, Mar 29, 2005 at 11:06:16AM -0500, Ed L Cashin wrote:
>> >
>> > NACK.  this changes devices nodes based on a compile-time option.  
>> 
>> I'm not sure I follow.  This configuration option sets the number of
>> partitions per device in the driver.  It doesn't create device nodes.
>> 
>> If the user has udev, then the device nodes are created correctly (on
>> Fedora Core 3), so that if the driver is configured with 1 partition
>> per device, the minor numbers for the disks are low.  
>> 
>> The folks I've talked to who aren't using udev but want one partition
>> per device already know that they have to re-create their device
>> files.
>
> It changes a kernel ABI, so people that have different config options
> set can't use the same userland.  It's a really big no-go.

I don't know if it matters now that we have udev.  When udev manages
the device nodes it all just works, so there's practically not much of
an issue.  The UUIDs in Software RAID and LVM make this even less of
an issue.  Besides, if they aren't using udev and are using a custom
kernel *and* using this configure option, then they're changing the
ABI on purpose for a practical benefit.

If you're saying that it's bad in principal, then that's another
story.  If that's what you mean, then it's a Linux policy issue, and
to follow convention I'd think that we'd need another major number.
That would be like the partitionable md devices, etc.

To me, the latter seems like the uglier way to go, and it would be
more permanent, while the configure option seems ugly but convenient
for this transitional stage, and could go away when high minor numbers
are well supported on the systems currently in use.

-- 
  Ed L Cashin <ecashin@coraid.com>

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Christoph Hellwig]

On Tue, Mar 29, 2005 at 11:48:48AM -0500, Ed L Cashin wrote:
> I don't know if it matters now that we have udev.  When udev manages
> the device nodes it all just works,

But most peopel still don't use udev.

> If you're saying that it's bad in principal, then that's another
> story.  If that's what you mean, then it's a Linux policy issue, and
> to follow convention I'd think that we'd need another major number.
> That would be like the partitionable md devices, etc.

Yes, it's a policy issue.  We don't do this weird config option anywhere
else.

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Ed L Cashin]

Christoph Hellwig <hch@infradead.org> writes:

> On Tue, Mar 29, 2005 at 11:48:48AM -0500, Ed L Cashin wrote:
>> I don't know if it matters now that we have udev.  When udev manages
>> the device nodes it all just works,
>
> But most peopel still don't use udev.
>
>> If you're saying that it's bad in principal, then that's another
>> story.  If that's what you mean, then it's a Linux policy issue, and
>> to follow convention I'd think that we'd need another major number.
>> That would be like the partitionable md devices, etc.
>
> Yes, it's a policy issue.  We don't do this weird config option anywhere
> else.

A couple support calls later, I think I've come around to your point
of view.  This patch isn't needed and may cause confusion.

Few aoe users really use partitions on their aoe disks, so I can make
the aoe driver have one minor number per disk as the default to avoid
the most common problems people encounter.

Then, aoe users who really need to partition their network disks can
use the partitionable md driver to "wrap" the aoe disk, like this:

  mdadm -B -l linear --force -n 1 --auto=mdp /dev/md_p0 /dev/etherd/e7.0
  fdisk /dev/md_p0


-- 
  Ed L Cashin <ecashin@coraid.com>

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Greg KH]

On Thu, Apr 07, 2005 at 02:28:59PM -0400, Ed L Cashin wrote:
> Christoph Hellwig <hch@infradead.org> writes:
> 
> > On Tue, Mar 29, 2005 at 11:48:48AM -0500, Ed L Cashin wrote:
> >> I don't know if it matters now that we have udev.  When udev manages
> >> the device nodes it all just works,
> >
> > But most peopel still don't use udev.
> >
> >> If you're saying that it's bad in principal, then that's another
> >> story.  If that's what you mean, then it's a Linux policy issue, and
> >> to follow convention I'd think that we'd need another major number.
> >> That would be like the partitionable md devices, etc.
> >
> > Yes, it's a policy issue.  We don't do this weird config option anywhere
> > else.
> 
> A couple support calls later, I think I've come around to your point
> of view.  This patch isn't needed and may cause confusion.
> 
> Few aoe users really use partitions on their aoe disks, so I can make
> the aoe driver have one minor number per disk as the default to avoid
> the most common problems people encounter.
> 
> Then, aoe users who really need to partition their network disks can
> use the partitionable md driver to "wrap" the aoe disk, like this:
> 
>   mdadm -B -l linear --force -n 1 --auto=mdp /dev/md_p0 /dev/etherd/e7.0
>   fdisk /dev/md_p0

So, which one of the aoe patches listed at:
	http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/driver/
do you want me to drop?  This one:
	http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/driver/aoe-AOE_PARTITIONS.patch
?
Or some other one too?

thanks,

greg k-h

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Ed L Cashin]

Greg KH <greg@kroah.com> writes:

...
> So, which one of the aoe patches listed at:
> 	http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/driver/
> do you want me to drop?  This one:
> 	http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/driver/aoe-AOE_PARTITIONS.patch
> ?
> Or some other one too?

Just aoe-AOE_PARTITIONS.patch, the seventh of the twelve, should be
dropped.

Then later I'll send a batch of patches that will include a change to
make aoe disks non-partitionable by default.

-- 
  Ed L Cashin <ecashin@coraid.com>

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Greg KH]

On Thu, Apr 07, 2005 at 02:56:39PM -0400, Ed L Cashin wrote:
> Greg KH <greg@kroah.com> writes:
> 
> ...
> > So, which one of the aoe patches listed at:
> > 	http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/driver/
> > do you want me to drop?  This one:
> > 	http://www.kernel.org/pub/linux/kernel/people/gregkh/gregkh-2.6/patches/driver/aoe-AOE_PARTITIONS.patch
> > ?
> > Or some other one too?
> 
> Just aoe-AOE_PARTITIONS.patch, the seventh of the twelve, should be
> dropped.

Ok, dropped.

> Then later I'll send a batch of patches that will include a change to
> make aoe disks non-partitionable by default.

That's fine.  Mind if I forward the other aoe patches in that directory
to Linus soon?

thanks,

greg k-h

Re: [PATCH 2.6.11] aoe [7/12]: support configuration of AOE_PARTITIONS from Kconfig

[Ed L Cashin]

Greg KH <greg@kroah.com> writes:

> On Thu, Apr 07, 2005 at 02:56:39PM -0400, Ed L Cashin wrote:
...
>> Just aoe-AOE_PARTITIONS.patch, the seventh of the twelve, should be
>> dropped.
>
> Ok, dropped.
>
>> Then later I'll send a batch of patches that will include a change to
>> make aoe disks non-partitionable by default.
>
> That's fine.  Mind if I forward the other aoe patches in that directory
> to Linus soon?

Go ahead.  I have a new batch of patches to send, but it looks like I
might not get to it for a few days.

-- 
  Ed L Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [8/12]: document env var for specifying number of partitions per dev

[ecashin]

document env var for specifying number of partitions per dev

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/Documentation/aoe/mkdevs.sh b/Documentation/aoe/mkdevs.sh
--- a/Documentation/aoe/mkdevs.sh	2005-03-07 17:37:14.000000000 -0500
+++ b/Documentation/aoe/mkdevs.sh	2005-03-10 12:19:56.000000000 -0500
@@ -5,6 +5,7 @@ n_partitions=${n_partitions:-16}
 
 if test "$#" != "1"; then
 	echo "Usage: sh `basename $0` {dir}" 1>&2
+	echo "       n_partitions=16 sh `basename $0` {dir}" 1>&2
 	exit 1
 fi
 dir=$1
diff -uprN a/Documentation/aoe/mkshelf.sh b/Documentation/aoe/mkshelf.sh
--- a/Documentation/aoe/mkshelf.sh	2005-03-07 17:37:14.000000000 -0500
+++ b/Documentation/aoe/mkshelf.sh	2005-03-10 12:19:56.000000000 -0500
@@ -2,6 +2,7 @@
 
 if test "$#" != "2"; then
 	echo "Usage: sh `basename $0` {dir} {shelfaddress}" 1>&2
+	echo "       n_partitions=16 sh `basename $0` {dir} {shelfaddress}" 1>&2
 	exit 1
 fi
 n_partitions=${n_partitions:-16}


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [9/12]: add note about the need for deadlock-free sk_buff allocation

[ecashin]

add note about the need for deadlock-free sk_buff allocation

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/Documentation/aoe/todo.txt b/Documentation/aoe/todo.txt
--- a/Documentation/aoe/todo.txt	1969-12-31 19:00:00.000000000 -0500
+++ b/Documentation/aoe/todo.txt	2005-03-10 12:19:57.000000000 -0500
@@ -0,0 +1,14 @@
+There is a potential for deadlock when allocating a struct sk_buff for
+data that needs to be written out to aoe storage.  If the data is
+being written from a dirty page in order to free that page, and if
+there are no other pages available, then deadlock may occur when a
+free page is needed for the sk_buff allocation.  This situation has
+not been observed, but it would be nice to eliminate any potential for
+deadlock under memory pressure.
+
+Because ATA over Ethernet is not fragmented by the kernel's IP code,
+the destructore member of the struct sk_buff is available to the aoe
+driver.  By using a mempool for allocating all but the first few
+sk_buffs, and by registering a destructor, we should be able to
+efficiently allocate sk_buffs without introducing any potential for
+deadlock.


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [10/12]: Randy Dunlap: avoid warnings on sparc64

[ecashin]

Andrew Morton has already merged this patch.

Randy Dunlap: avoid warnings on sparc64

Signed-off-by: Randy Dunlap <rddunlap@osdl.org>
Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c
--- a/drivers/block/aoe/aoeblk.c	2005-03-07 17:37:14.000000000 -0500
+++ b/drivers/block/aoe/aoeblk.c	2005-03-10 12:20:00.000000000 -0500
@@ -28,7 +28,8 @@ static ssize_t aoedisk_show_mac(struct g
 {
 	struct aoedev *d = disk->private_data;
 
-	return snprintf(page, PAGE_SIZE, "%012llx\n", mac_addr(d->addr));
+	return snprintf(page, PAGE_SIZE, "%012llx\n",
+			(unsigned long long)mac_addr(d->addr));
 }
 static ssize_t aoedisk_show_netif(struct gendisk * disk, char *page)
 {
@@ -241,7 +242,8 @@ aoeblk_gdalloc(void *vp)
 	aoedisk_add_sysfs(d);
 	
 	printk(KERN_INFO "aoe: %012llx e%lu.%lu v%04x has %llu "
-		"sectors\n", mac_addr(d->addr), d->aoemajor, d->aoeminor,
+		"sectors\n", (unsigned long long)mac_addr(d->addr),
+		d->aoemajor, d->aoeminor,
 		d->fw_ver, (long long)d->ssize);
 }
 


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [11/12]: add support for disk statistics

[ecashin]

add support for disk statistics

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
--- a/drivers/block/aoe/aoe.h	2005-03-10 12:19:54.000000000 -0500
+++ b/drivers/block/aoe/aoe.h	2005-03-10 12:20:02.000000000 -0500
@@ -91,6 +91,7 @@ enum {
 
 struct buf {
 	struct list_head bufs;
+	ulong start_time;	/* for disk stats */
 	ulong flags;
 	ulong nframesout;
 	char *bufaddr;
diff -uprN a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c
--- a/drivers/block/aoe/aoeblk.c	2005-03-10 12:20:00.000000000 -0500
+++ b/drivers/block/aoe/aoeblk.c	2005-03-10 12:20:02.000000000 -0500
@@ -125,6 +125,7 @@ aoeblk_make_request(request_queue_t *q, 
 	}
 	memset(buf, 0, sizeof(*buf));
 	INIT_LIST_HEAD(&buf->bufs);
+	buf->start_time = jiffies;
 	buf->bio = bio;
 	buf->resid = bio->bi_size;
 	buf->sector = bio->bi_sector;
diff -uprN a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
--- a/drivers/block/aoe/aoecmd.c	2005-03-10 12:19:27.000000000 -0500
+++ b/drivers/block/aoe/aoecmd.c	2005-03-10 12:20:02.000000000 -0500
@@ -456,6 +456,20 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 	if (buf) {
 		buf->nframesout -= 1;
 		if (buf->nframesout == 0 && buf->resid == 0) {
+			unsigned long duration = jiffies - buf->start_time;
+			unsigned long n_sect = buf->bio->bi_size >> 9;
+			struct gendisk *disk = d->gd;
+
+			if (bio_data_dir(buf->bio) == WRITE) {
+				disk_stat_inc(disk, writes);
+				disk_stat_add(disk, write_ticks, duration);
+				disk_stat_add(disk, write_sectors, n_sect);
+			} else {
+				disk_stat_inc(disk, reads);
+				disk_stat_add(disk, read_ticks, duration);
+				disk_stat_add(disk, read_sectors, n_sect);
+			}
+			disk_stat_add(disk, io_ticks, duration);
 			n = (buf->flags & BUFFL_FAIL) ? -EIO : 0;
 			bio_endio(buf->bio, buf->bio->bi_size, n);
 			mempool_free(buf, d->bufpool);


-- 
  Ed L. Cashin <ecashin@coraid.com>

[PATCH 2.6.11] aoe [12/12]: send outgoing packets in order

[ecashin]

I can't use list.h, since sk_buff doesn't have a list_head but instead
has two struct sk_buff pointers, and I want to avoid any extra memory
allocation.

send outgoing packets in order

Signed-off-by: Ed L. Cashin <ecashin@coraid.com>

diff -uprN a/drivers/block/aoe/aoe.h b/drivers/block/aoe/aoe.h
--- a/drivers/block/aoe/aoe.h	2005-03-10 12:20:02.000000000 -0500
+++ b/drivers/block/aoe/aoe.h	2005-03-10 12:20:04.000000000 -0500
@@ -132,7 +132,8 @@ struct aoedev {
 	struct timer_list timer;
 	spinlock_t lock;
 	struct net_device *ifp;	/* interface ed is attached to */
-	struct sk_buff *skblist;/* packets needing to be sent */
+	struct sk_buff *sendq_hd; /* packets needing to be sent, list head */
+	struct sk_buff *sendq_tl;
 	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 */
diff -uprN a/drivers/block/aoe/aoeblk.c b/drivers/block/aoe/aoeblk.c
--- a/drivers/block/aoe/aoeblk.c	2005-03-10 12:20:02.000000000 -0500
+++ b/drivers/block/aoe/aoeblk.c	2005-03-10 12:20:04.000000000 -0500
@@ -147,8 +147,8 @@ aoeblk_make_request(request_queue_t *q, 
 	list_add_tail(&buf->bufs, &d->bufq);
 	aoecmd_work(d);
 
-	sl = d->skblist;
-	d->skblist = NULL;
+	sl = d->sendq_hd;
+	d->sendq_hd = d->sendq_tl = NULL;
 
 	spin_unlock_irqrestore(&d->lock, flags);
 
diff -uprN a/drivers/block/aoe/aoecmd.c b/drivers/block/aoe/aoecmd.c
--- a/drivers/block/aoe/aoecmd.c	2005-03-10 12:20:02.000000000 -0500
+++ b/drivers/block/aoe/aoecmd.c	2005-03-10 12:20:04.000000000 -0500
@@ -178,8 +178,12 @@ aoecmd_ata_rw(struct aoedev *d, struct f
 
 	skb = skb_prepare(d, f);
 	if (skb) {
-		skb->next = d->skblist;
-		d->skblist = skb;
+		skb->next = NULL;
+		if (d->sendq_hd)
+			d->sendq_tl->next = skb;
+		else
+			d->sendq_hd = skb;
+		d->sendq_tl = skb;
 	}
 }
 
@@ -227,8 +231,12 @@ rexmit(struct aoedev *d, struct frame *f
 
 	skb = skb_prepare(d, f);
 	if (skb) {
-		skb->next = d->skblist;
-		d->skblist = skb;
+		skb->next = NULL;
+		if (d->sendq_hd)
+			d->sendq_tl->next = skb;
+		else
+			d->sendq_hd = skb;
+		d->sendq_tl = skb;
 	}
 }
 
@@ -280,8 +288,8 @@ tdie:		spin_unlock_irqrestore(&d->lock, 
 		}
 	}
 
-	sl = d->skblist;
-	d->skblist = NULL;
+	sl = d->sendq_hd;
+	d->sendq_hd = d->sendq_tl = NULL;
 	if (sl) {
 		n = d->rttavg <<= 1;
 		if (n > MAXTIMER)
@@ -481,8 +489,8 @@ aoecmd_ata_rsp(struct sk_buff *skb)
 
 	aoecmd_work(d);
 
-	sl = d->skblist;
-	d->skblist = NULL;
+	sl = d->sendq_hd;
+	d->sendq_hd = d->sendq_tl = NULL;
 
 	spin_unlock_irqrestore(&d->lock, flags);
 
@@ -531,7 +539,7 @@ aoecmd_cfg(ushort aoemajor, unsigned cha
  
 /*
  * Since we only call this in one place (and it only prepares one frame)
- * we just return the skb.  Usually we'd chain it up to the d->skblist.
+ * we just return the skb.  Usually we'd chain it up to the aoedev sendq.
  */
 static struct sk_buff *
 aoecmd_ata_id(struct aoedev *d)


-- 
  Ed L. Cashin <ecashin@coraid.com>