* [PATCH] dev: use name hash for dev_seq_ops
From: Mihai Maruseac @ 2011-10-20 8:01 UTC (permalink / raw)
To: davem, shemminger, eric.dumazet
Cc: mirq-linux, therbert, jpirko, netdev, linux-kernel, dbaluta,
Mihai Maruseac
Instead of using the dev->next chain and trying to resync at each call to
dev_seq_start, use the name hash, keeping the bucket and the offset in
seq->private field.
The bucket number and the offset in the bucket are kept in a single unsigned
int value: pos = bucket << (32 - NETDEV_HASHBITS) + offset. When we reach the
end of the hash, both the bucket and the offset reach 0, so we need to test for
this in dev_seq_start to prevent endless loops.
Tests revealed the following results for ifconfig > /dev/null
* 1000 interfaces:
* 0.114s without patch
* 0.089s with patch
* 3000 interfaces:
* 0.489s without patch
* 0.110s with patch
* 5000 interfaces:
* 1.363s without patch
* 0.250s with patch
* 128000 interfaces (other setup):
* ~100s without patch
* ~30s with patch
Signed-off-by: Mihai Maruseac <mmaruseac@ixiacom.com>
---
net/core/dev.c | 83 +++++++++++++++++++++++++++++++++++++++++++++----------
1 files changed, 68 insertions(+), 15 deletions(-)
diff --git a/net/core/dev.c b/net/core/dev.c
index 70ecb86..103eb28 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -4041,6 +4041,59 @@ static int dev_ifconf(struct net *net, char __user *arg)
}
#ifdef CONFIG_PROC_FS
+
+struct dev_iter_state {
+ struct seq_net_private p;
+ unsigned int pos; /* bucket << 24 + offset */
+};
+
+#define BUCKET_SPACE (32 - NETDEV_HASHBITS)
+#define get_bucket(x) ((x) >> BUCKET_SPACE)
+#define get_offset(x) ((x) & ((1 << BUCKET_SPACE) - 1))
+#define set_bucket_offset(b, o) ((b) << BUCKET_SPACE | (o))
+
+static inline struct net_device *dev_from_same_bucket(struct seq_file *seq)
+{
+ struct dev_iter_state *state = seq->private;
+ struct net *net = seq_file_net(seq);
+ struct net_device *dev;
+ struct hlist_node *p;
+ struct hlist_head *h;
+ unsigned int count, bucket, offset;
+
+ bucket = get_bucket(state->pos);
+ offset = get_offset(state->pos);
+ h = &net->dev_name_head[bucket];
+ count = 0;
+ hlist_for_each_entry_rcu(dev, p, h, name_hlist) {
+ if (count++ == offset) {
+ state->pos = set_bucket_offset(bucket, count);
+ return dev;
+ }
+ }
+
+ return NULL;
+}
+
+static inline struct net_device *dev_from_new_bucket(struct seq_file *seq)
+{
+ struct dev_iter_state *state = seq->private;
+ struct net_device *dev;
+ unsigned int bucket;
+
+ bucket = get_bucket(state->pos);
+ do {
+ dev = dev_from_same_bucket(seq);
+ if (dev)
+ return dev;
+
+ bucket++;
+ state->pos = set_bucket_offset(bucket, 0);
+ } while (bucket < NETDEV_HASHENTRIES);
+
+ return NULL;
+}
+
/*
* This is invoked by the /proc filesystem handler to display a device
* in detail.
@@ -4048,33 +4101,33 @@ static int dev_ifconf(struct net *net, char __user *arg)
void *dev_seq_start(struct seq_file *seq, loff_t *pos)
__acquires(RCU)
{
- struct net *net = seq_file_net(seq);
- loff_t off;
- struct net_device *dev;
+ struct dev_iter_state *state = seq->private;
rcu_read_lock();
if (!*pos)
return SEQ_START_TOKEN;
- off = 1;
- for_each_netdev_rcu(net, dev)
- if (off++ == *pos)
- return dev;
+ /* check for end of the hash */
+ if (state->pos == 0 && *pos > 1)
+ return NULL;
- return NULL;
+ return dev_from_new_bucket(seq);
}
void *dev_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
- struct net_device *dev = v;
+ struct net_device *dev;
+
+ ++*pos;
if (v == SEQ_START_TOKEN)
- dev = first_net_device_rcu(seq_file_net(seq));
- else
- dev = next_net_device_rcu(dev);
+ return dev_from_new_bucket(seq);
- ++*pos;
- return dev;
+ dev = dev_from_same_bucket(seq);
+ if (dev)
+ return dev;
+
+ return dev_from_new_bucket(seq);
}
void dev_seq_stop(struct seq_file *seq, void *v)
@@ -4173,7 +4226,7 @@ static const struct seq_operations dev_seq_ops = {
static int dev_seq_open(struct inode *inode, struct file *file)
{
return seq_open_net(inode, file, &dev_seq_ops,
- sizeof(struct seq_net_private));
+ sizeof(struct dev_iter_state));
}
static const struct file_operations dev_seq_fops = {
--
1.7.4.1
^ permalink raw reply related
* [PATCH] virtio_net: Clean up set_skb_frag()
From: Krishna Kumar @ 2011-10-20 8:17 UTC (permalink / raw)
To: rusty, mst; +Cc: netdev, linux-kernel, davem, Krishna Kumar, virtualization
Remove manual initialization in set_skb_frag, and instead
use __skb_fill_page_desc() to do the same. Patch tested
on net-next.
Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
---
drivers/net/virtio_net.c | 13 +++++--------
1 file changed, 5 insertions(+), 8 deletions(-)
diff -ruNp org/drivers/net/virtio_net.c new/drivers/net/virtio_net.c
--- org/drivers/net/virtio_net.c 2011-10-20 08:34:31.000000000 +0530
+++ new/drivers/net/virtio_net.c 2011-10-20 08:43:57.000000000 +0530
@@ -143,18 +143,15 @@ static void skb_xmit_done(struct virtque
static void set_skb_frag(struct sk_buff *skb, struct page *page,
unsigned int offset, unsigned int *len)
{
+ int size = min((unsigned)PAGE_SIZE - offset, *len);
int i = skb_shinfo(skb)->nr_frags;
- skb_frag_t *f;
- f = &skb_shinfo(skb)->frags[i];
- skb_frag_size_set(f, min((unsigned)PAGE_SIZE - offset, *len));
- f->page_offset = offset;
- __skb_frag_set_page(f, page);
+ __skb_fill_page_desc(skb, i, page, offset, size);
- skb->data_len += skb_frag_size(f);
- skb->len += skb_frag_size(f);
+ skb->data_len += size;
+ skb->len += size;
skb_shinfo(skb)->nr_frags++;
- *len -= skb_frag_size(f);
+ *len -= size;
}
static struct sk_buff *page_to_skb(struct virtnet_info *vi,
^ permalink raw reply
* Re: [patch net-next]alx: Atheros AR8131/AR8151/AR8152/AR8161 Ethernet driver
From: David Miller @ 2011-10-20 8:45 UTC (permalink / raw)
To: cloud.ren; +Cc: Luis.Rodriguez, netdev, linux-kernel
In-Reply-To: <1319093184-19696-1-git-send-email-cloud.ren@Atheros.com>
From: <cloud.ren@Atheros.com>
Date: Thu, 20 Oct 2011 14:46:24 +0800
> +#define __far
So much unused crap left in these header files, get rid of this stuff.
+#define ALX_HW_WARN(_fmt, _args...) \
+ ALX_HW_PRINTA(WARNING, _fmt, ## _args)
+
+#define ALX_HW_INFO(_fmt, _args...) \
+ ALX_HW_PRINTA(INFO, _fmt, ## _args)
+
+#define ALX_HW_DBG(_fmt, _args...) \
+ ALX_HW_PRINTA(DEBUG, _fmt, ## _args)
+
We told you to get rid of your customized debug logging interfaces,
yet all of this stuff is still there.
+/* delay function */
+#define US_DELAY(_hw, _n) __US_DELAY(_n)
+#define MS_DELAY(_hw, _n) __MS_DELAY(_n)
+#define __US_DELAY(_n) udelay(_n)
+#define __MS_DELAY(_n) mdelay(_n)
Useless wrappers for standard kernel interfaces, kill this.
+#define DEBUG_INFO(_a, _b)
+#define DEBUG_INFOS(_a, _b)
Again we told you to get rid of this stuff.
I suspect it's going to take may rounds of feedback before this driver
is anywhere near ready for inclusion.
Please just submit it to staging and let it cook there for a couple
weeks in the interests of our sanity.
^ permalink raw reply
* Re: [PATCH] virtio_net: Clean up set_skb_frag()
From: Eric Dumazet @ 2011-10-20 8:46 UTC (permalink / raw)
To: Krishna Kumar; +Cc: rusty, mst, netdev, linux-kernel, davem, virtualization
In-Reply-To: <20111020081727.509.83122.sendpatchset@krkumar2.in.ibm.com>
Le jeudi 20 octobre 2011 à 13:47 +0530, Krishna Kumar a écrit :
> Remove manual initialization in set_skb_frag, and instead
> use __skb_fill_page_desc() to do the same. Patch tested
> on net-next.
>
> Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
> ---
> drivers/net/virtio_net.c | 13 +++++--------
> 1 file changed, 5 insertions(+), 8 deletions(-)
>
> diff -ruNp org/drivers/net/virtio_net.c new/drivers/net/virtio_net.c
> --- org/drivers/net/virtio_net.c 2011-10-20 08:34:31.000000000 +0530
> +++ new/drivers/net/virtio_net.c 2011-10-20 08:43:57.000000000 +0530
> @@ -143,18 +143,15 @@ static void skb_xmit_done(struct virtque
> static void set_skb_frag(struct sk_buff *skb, struct page *page,
> unsigned int offset, unsigned int *len)
> {
> + int size = min((unsigned)PAGE_SIZE - offset, *len);
> int i = skb_shinfo(skb)->nr_frags;
> - skb_frag_t *f;
>
> - f = &skb_shinfo(skb)->frags[i];
> - skb_frag_size_set(f, min((unsigned)PAGE_SIZE - offset, *len));
> - f->page_offset = offset;
> - __skb_frag_set_page(f, page);
> + __skb_fill_page_desc(skb, i, page, offset, size);
>
> - skb->data_len += skb_frag_size(f);
> - skb->len += skb_frag_size(f);
> + skb->data_len += size;
> + skb->len += size;
> skb_shinfo(skb)->nr_frags++;
> - *len -= skb_frag_size(f);
> + *len -= size;
> }
>
> static struct sk_buff *page_to_skb(struct virtnet_info *vi,
>
> --
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Seems fine to me. There is a truesize patch for this driver, I'll send
it once your patch is in.
^ permalink raw reply
* Re: [patch net-next]alx: Atheros AR8131/AR8151/AR8152/AR8161 Ethernet driver
From: David Miller @ 2011-10-20 8:48 UTC (permalink / raw)
To: cjren; +Cc: romieu, rodrigue, netdev, linux-kernel
In-Reply-To: <6349D7A510622448B1BA0967850A8438011CC12D@nasanexd02d.na.qualcomm.com>
From: "Ren, Cloud" <cjren@qca.qualcomm.com>
Date: Thu, 20 Oct 2011 07:29:27 +0000
> Thanks for your information.
>
> Drivers on different OSs share alf_hw.c, alf_hw.h, alc_hw.c and alc_hw.h
>
> So it looks some of codes are duplicate, because other drivers need them.
You cannot do this, you must get rid of such compatability wrappers
and unused code.
^ permalink raw reply
* Re: [v3 PATCH 0/2] NETFILTER new target module, HMARK
From: Pablo Neira Ayuso @ 2011-10-20 8:54 UTC (permalink / raw)
To: Hans Schillstrom; +Cc: kaber, jengelh, netfilter-devel, netdev, hans
In-Reply-To: <1318532530-29446-1-git-send-email-hans.schillstrom@ericsson.com>
Hi Hans,
On Thu, Oct 13, 2011 at 09:02:08PM +0200, Hans Schillstrom wrote:
> The target allows you to create rules in the "raw" and "mangle" tables
> which alter the netfilter mark (nfmark) field within a given range.
> First a 32 bit hash value is generated then modulus by <limit> and
> finally an offset is added before it's written to nfmark.
> Prior to routing, the nfmark can influence the routing method (see
> "Use netfilter MARK value as routing key") and can also be used by
> other subsystems to change their behaviour.
>
> The mark match can also be used to match nfmark produced by this module.
> See the kernel module for more info.
>
> REVISION
> Version 3
> Handling of SCTP for IPv6 added.
>
> Version 2
> NAT Added for IPv4
> IPv6 ICMP handling enhanced.
> Usage example added
>
> Version 1
> Initial RFC
>
>
> We (Ericsson) use hmark in-front of ipvs as a pre-loadbalancer and
> handles up to 70 ipvs running in parallel in clusters.
> However hmark is not restricted to run infront of IPVS it can also be used as
> "poor mans" load balancer.
> With this version is also NAT supported as an option, with very high flows
> you might not want to use conntrack.
>
> The idea is to generate a direction independent fw mark range to use as input to
> the routing (i.e. ip rule add fwmark ...).
> Pretty straight forward and simple.
>
>
> Example:
> App Server (Real Server)
>
> +---------+
> -->| Service |
> Gateway A +---------+
> /
> +----------+ / +----+ +---------+
> --- if -A---| selector |----> |ipvs| --->| Service |
> +----------+ \ +----+ +---------+
> \
> +----+ +---------+
> |ipvs| -->| Service |
> +----+ +---------+
> Gateway C
> +----------+ / +----+
> --- if-B ---| selector | ---> |ipvs|
> +----------+ \ +----+ +---------+
> | Service |
> +---------+
> /
> +----------+ / +----+ ..
> --- if-B ---| selector | ---> |ipvs| +---------+
> +----------+ \ +----+ | Service |
> \ +---------+
> #
> # Example with four ipvs loadbalancers
> #
> iptables -t mangle -I PREROUTING -d $IPADDR -j HMARK --hmark-mod 4 --hmark-offs 100
I think you can replace this rule by:
iptables -t mangle -I PREROUTING -d $IPADDR -m cluster \
--cluster-total-nodes 4 --cluster-local-node 1
--cluster-hash-seed 0xdeadbeef -j MARK --set-mark 100
iptables -t mangle -I PREROUTING -d $IPADDR -m cluster \
--cluster-total-nodes 4 --cluster-local-node 2
--cluster-hash-seed 0xdeadbeef -j MARK --set-mark 101
iptables -t mangle -I PREROUTING -d $IPADDR -m cluster \
--cluster-total-nodes 4 --cluster-local-node 3
--cluster-hash-seed 0xdeadbeef -j MARK --set-mark 103
iptables -t mangle -I PREROUTING -d $IPADDR -m cluster \
--cluster-total-nodes 4 --cluster-local-node 4
--cluster-hash-seed 0xdeadbeef -j MARK --set-mark 104
The hashing is done by the cluster match, which is currently based on
the source address.
This match currently depends on the connection tracking system, so you
could save the ctmark with CONNMARK. Thus, you only has to hash the
first packet of the flow, instead of hashing every single packet.
> ip rule add fwmark 100 table 100
> ip rule add fwmark 101 table 101
> ip rule add fwmark 102 table 102
> ip rule add fwmark 103 table 103
>
> ip ro ad table 100 default via x.y.z.1 dev bond1
> ip ro ad table 101 default via x.y.z.2 dev bond1
> ip ro ad table 102 default via x.y.z.3 dev bond1
> ip ro ad table 103 default via x.y.z.4 dev bond1
>
> If conntrack doesn't handle the return path,
> do the oposite with HMARK and send it back right to ipvs.
>
> Another exmaple of usage could be if you have cluster originated connections
> and want to spread the connections over a number of interfaces
> (NAT will complpicate things for you in this case)
>
>
>
> \ Blade 1
> \ +----------+ +---------+
> <-- | selector | <--- | Service |
> / +----------+ +---------+
> /
> +------+
> -- | Gw-A | \ Blade 2
> +------+ \ +----------+ +---------+
> +------+ <-- | selector | <--- | Service |
> -- | Gw-B | / +----------+ +---------+
> +------+ /
> +------+
> -- | Gw-C | \
> +------+ \ +----------+ +---------+
> <-- | selector | <--- | Service |
> / +----------+ +---------+
> /
>
> \ Blande -n
> \ +----------+ +---------+
> <-- | selector | <--- | Service |
> / +----------+ +---------+
> /
Unless I'm missing something, I think this can be done with the
cluster match as well.
^ permalink raw reply
* Re: [PATCH] virtio_net: Clean up set_skb_frag()
From: David Miller @ 2011-10-20 8:54 UTC (permalink / raw)
To: eric.dumazet; +Cc: krkumar2, rusty, mst, netdev, linux-kernel, virtualization
In-Reply-To: <1319100362.3781.3.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC>
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Thu, 20 Oct 2011 10:46:02 +0200
> Le jeudi 20 octobre 2011 à 13:47 +0530, Krishna Kumar a écrit :
>> Remove manual initialization in set_skb_frag, and instead
>> use __skb_fill_page_desc() to do the same. Patch tested
>> on net-next.
>>
>> Signed-off-by: Krishna Kumar <krkumar2@in.ibm.com>
...
> Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Applied, thanks.
^ permalink raw reply
* Re: [PATCH V2 net-next] mlx4_en: fix skb truesize underestimation
From: David Miller @ 2011-10-20 8:55 UTC (permalink / raw)
To: eric.dumazet; +Cc: netdev, yevgenyp
In-Reply-To: <1319086192.8416.42.camel@edumazet-laptop>
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Thu, 20 Oct 2011 06:49:52 +0200
> skb->truesize must account for allocated memory, not the used part of
> it. Doing this work is important to avoid unexpected OOM situations.
>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> CC: Yevgeny Petrilin <yevgenyp@mellanox.co.il>
> ---
> V2: respin after recent patches
Applied, thanks Eric.
^ permalink raw reply
* [PATCH net-next] bnx2x: fix skb truesize underestimation
From: Eric Dumazet @ 2011-10-20 9:00 UTC (permalink / raw)
To: David Miller; +Cc: netdev, Eilon Greenstein
bnx2x allocates a full page per fragment.
We must account in skb->truesize, the size of the fragment, not the used
part of it.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Eilon Greenstein <eilong@broadcom.com>
---
drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
index dd8ee56..580b44e 100644
--- a/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
+++ b/drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c
@@ -454,7 +454,7 @@ static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
skb_fill_page_desc(skb, j, old_rx_pg.page, 0, frag_len);
skb->data_len += frag_len;
- skb->truesize += frag_len;
+ skb->truesize += SGE_PAGE_SIZE * PAGES_PER_SGE;
skb->len += frag_len;
frag_size -= frag_len;
^ permalink raw reply related
* Re: [patch net-next]alx: Atheros AR8131/AR8151/AR8152/AR8161 Ethernet driver
From: Joe Perches @ 2011-10-20 9:00 UTC (permalink / raw)
To: David Miller; +Cc: cloud.ren, Luis.Rodriguez, netdev, linux-kernel
In-Reply-To: <20111020.044541.970282389722164761.davem@davemloft.net>
On Thu, 2011-10-20 at 04:45 -0400, David Miller wrote:
> From: <cloud.ren@Atheros.com>
[]
> +#define ALX_HW_WARN(_fmt, _args...) \
> + ALX_HW_PRINTA(WARNING, _fmt, ## _args)
> +
> +#define ALX_HW_INFO(_fmt, _args...) \
> + ALX_HW_PRINTA(INFO, _fmt, ## _args)
> +
> +#define ALX_HW_DBG(_fmt, _args...) \
> + ALX_HW_PRINTA(DEBUG, _fmt, ## _args)
> +
I've just done patches for these
and for ALX_PRINTA and ALX_PRINTB.
I'll send them directly to Cloud.
> Please just submit it to staging and let it cook there for a couple
> weeks in the interests of our sanity.
That's a good plan.
^ permalink raw reply
* [PATCH 0/6] skb fragment API: convert network drivers (part V, take 2)
From: Ian Campbell @ 2011-10-20 9:01 UTC (permalink / raw)
To: netdev@vger.kernel.org; +Cc: linux-scsi@vger.kernel.org, linux-mm@kvack.org
The following series is the second attempt to convert a fifth (and
hopefully final) batch of network drivers to the SKB pages fragment API
introduced in 131ea6675c76.
There are four drivers here (mlx4, cxgb4, cxgb4vf and cxgbi) which used
skb_frag_t as part of their internal datastructures which meant that
they are impacted by changes to that type more than most drivers. To
break this dependency I added a "struct page_frag" (struct page + offset
+ len) and converted them to use it. These conversions are a little less
trivial than most of the preceding ones and I have only been able to
compile test them.
The struct page_frag addition has been acked by Andrew Morton to go
through the net tree. (Andrew, I took you "yes please" as an Acked-by. I
hope that's ok).
The final patch here wraps the page member of skb_frag_t in a structure,
this is a precursor to adding the destructor here (those patches need a
little more work, arising from comments made at LPC, I'll post regarding
those shortly). This should help ensure that no direct uses of the page
get introduced in the meantime.
In the previous posting of this series I ran an allmodconfig build on a
boatload architectures[2] on a baseline of the then current
net-next/master (88c5100c28b0) and with that series. Although the
baseline didn't build for most architectures I used "make -k" and
confirmed that this series added no new warnings or errors. For this
iteration I have just rebuilt things which changed in the interval
88c5100c28b0..a0bec1cd8f7a (current net-next/master) on amd64 and
eyeballed the diff for new uses of frag->page (I saw none).
This is part of my series to enable visibility into SKB paged fragment's
lifecycles, [0] contains some more background and rationale but
basically the completed series will allow entities which inject pages
into the networking stack to receive a notification when the stack has
really finished with those pages (i.e. including retransmissions,
clones, pull-ups etc) and not just when the original skb is finished
with, which is beneficial to many subsystems which wish to inject pages
into the network stack without giving up full ownership of those page's
lifecycle. It implements something broadly along the lines of what was
described in [1].
Cheers,
Ian.
[0] http://marc.info/?l=linux-netdev&m=131072801125521&w=2
[1] http://marc.info/?l=linux-netdev&m=130925719513084&w=2
[2] arm amd64 blackfin cris i386 ia64 m68k mips64 mips powerpc64 powerpc
s390x sh4 sparc64 sparc xtensa
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply
* [PATCH 1/6] mm: add a "struct page_frag" type containing a page, offset and length
From: Ian Campbell @ 2011-10-20 9:01 UTC (permalink / raw)
To: netdev@vger.kernel.org
Cc: Ian Campbell, Christoph Hellwig, David Miller, linux-mm,
linux-kernel
In-Reply-To: <1319101275.3385.129.camel@zakaz.uk.xensource.com>
A few network drivers currently use skb_frag_struct for this purpose but I have
patches which add additional fields and semantics there which these other uses
do not want.
A structure for reference sub-page regions seems like a generally useful thing
so do so instead of adding a network subsystem specific structure.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Acked-by: Jens Axboe <jaxboe@fusionio.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: David Miller <davem@davemloft.net>
Cc: linux-mm@kvack.org
Cc: linux-kernel@vger.kernel.org
[since v1: s/struct subpage/struct page_frag/ on advice from Christoph]
[since v2: s/page_offset/offset/ on advice from Andrew]
---
include/linux/mm_types.h | 11 +++++++++++
1 files changed, 11 insertions(+), 0 deletions(-)
diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
index 774b895..29971a5 100644
--- a/include/linux/mm_types.h
+++ b/include/linux/mm_types.h
@@ -135,6 +135,17 @@ struct page {
#endif
;
+struct page_frag {
+ struct page *page;
+#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
+ __u32 offset;
+ __u32 size;
+#else
+ __u16 offset;
+ __u16 size;
+#endif
+};
+
typedef unsigned long __nocast vm_flags_t;
/*
--
1.7.2.5
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply related
* [PATCH 5/6] cxgbi: convert to SKB paged frag API.
From: Ian Campbell @ 2011-10-20 9:01 UTC (permalink / raw)
To: netdev@vger.kernel.org
Cc: Ian Campbell, James E.J. Bottomley, David S. Miller,
Mike Christie, James Bottomley, Karen Xie, linux-scsi, netdev
In-Reply-To: <1319101275.3385.129.camel@zakaz.uk.xensource.com>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: "James E.J. Bottomley" <JBottomley@parallels.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: James Bottomley <James.Bottomley@suse.de>
Cc: Karen Xie <kxie@chelsio.com>
Cc: linux-scsi@vger.kernel.org
Cc: netdev@vger.kernel.org
---
drivers/scsi/cxgbi/libcxgbi.c | 28 +++++++++++++++-------------
drivers/scsi/cxgbi/libcxgbi.h | 2 +-
2 files changed, 16 insertions(+), 14 deletions(-)
diff --git a/drivers/scsi/cxgbi/libcxgbi.c b/drivers/scsi/cxgbi/libcxgbi.c
index be69da3..1c1329b 100644
--- a/drivers/scsi/cxgbi/libcxgbi.c
+++ b/drivers/scsi/cxgbi/libcxgbi.c
@@ -1787,7 +1787,7 @@ static int sgl_seek_offset(struct scatterlist *sgl, unsigned int sgcnt,
}
static int sgl_read_to_frags(struct scatterlist *sg, unsigned int sgoffset,
- unsigned int dlen, skb_frag_t *frags,
+ unsigned int dlen, struct page_frag *frags,
int frag_max)
{
unsigned int datalen = dlen;
@@ -1814,8 +1814,8 @@ static int sgl_read_to_frags(struct scatterlist *sg, unsigned int sgoffset,
copy = min(datalen, sglen);
if (i && page == frags[i - 1].page &&
sgoffset + sg->offset ==
- frags[i - 1].page_offset + skb_frag_size(&frags[i - 1])) {
- skb_frag_size_add(&frags[i - 1], copy);
+ frags[i - 1].offset + frags[i - 1].size) {
+ frags[i - 1].size += copy;
} else {
if (i >= frag_max) {
pr_warn("too many pages %u, dlen %u.\n",
@@ -1824,8 +1824,8 @@ static int sgl_read_to_frags(struct scatterlist *sg, unsigned int sgoffset,
}
frags[i].page = page;
- frags[i].page_offset = sg->offset + sgoffset;
- skb_frag_size_set(&frags[i], copy);
+ frags[i].offset = sg->offset + sgoffset;
+ frags[i].size = copy;
i++;
}
datalen -= copy;
@@ -1944,15 +1944,15 @@ int cxgbi_conn_init_pdu(struct iscsi_task *task, unsigned int offset,
if (tdata->nr_frags > MAX_SKB_FRAGS ||
(padlen && tdata->nr_frags == MAX_SKB_FRAGS)) {
char *dst = skb->data + task->hdr_len;
- skb_frag_t *frag = tdata->frags;
+ struct page_frag *frag = tdata->frags;
/* data fits in the skb's headroom */
for (i = 0; i < tdata->nr_frags; i++, frag++) {
char *src = kmap_atomic(frag->page,
KM_SOFTIRQ0);
- memcpy(dst, src+frag->page_offset, skb_frag_size(frag));
- dst += skb_frag_size(frag);
+ memcpy(dst, src+frag->offset, frag->size);
+ dst += frag->size;
kunmap_atomic(src, KM_SOFTIRQ0);
}
if (padlen) {
@@ -1962,11 +1962,13 @@ int cxgbi_conn_init_pdu(struct iscsi_task *task, unsigned int offset,
skb_put(skb, count + padlen);
} else {
/* data fit into frag_list */
- for (i = 0; i < tdata->nr_frags; i++)
- get_page(tdata->frags[i].page);
-
- memcpy(skb_shinfo(skb)->frags, tdata->frags,
- sizeof(skb_frag_t) * tdata->nr_frags);
+ for (i = 0; i < tdata->nr_frags; i++) {
+ __skb_fill_page_desc(skb, i,
+ tdata->frags[i].page,
+ tdata->frags[i].offset,
+ tdata->frags[i].size);
+ skb_frag_ref(skb, i);
+ }
skb_shinfo(skb)->nr_frags = tdata->nr_frags;
skb->len += count;
skb->data_len += count;
diff --git a/drivers/scsi/cxgbi/libcxgbi.h b/drivers/scsi/cxgbi/libcxgbi.h
index 9267844..3a25b11 100644
--- a/drivers/scsi/cxgbi/libcxgbi.h
+++ b/drivers/scsi/cxgbi/libcxgbi.h
@@ -574,7 +574,7 @@ struct cxgbi_endpoint {
#define MAX_PDU_FRAGS ((ULP2_MAX_PDU_PAYLOAD + 512 - 1) / 512)
struct cxgbi_task_data {
unsigned short nr_frags;
- skb_frag_t frags[MAX_PDU_FRAGS];
+ struct page_frag frags[MAX_PDU_FRAGS];
struct sk_buff *skb;
unsigned int offset;
unsigned int count;
--
1.7.2.5
^ permalink raw reply related
* [PATCH 2/6] mlx4: convert to SKB paged frag API.
From: Ian Campbell @ 2011-10-20 9:01 UTC (permalink / raw)
To: netdev@vger.kernel.org; +Cc: Ian Campbell, netdev
In-Reply-To: <1319101275.3385.129.camel@zakaz.uk.xensource.com>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: netdev@vger.kernel.org
---
drivers/net/ethernet/mellanox/mlx4/en_rx.c | 32 ++++++++++++++--------------
drivers/net/ethernet/mellanox/mlx4/en_tx.c | 20 +++--------------
2 files changed, 20 insertions(+), 32 deletions(-)
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_rx.c b/drivers/net/ethernet/mellanox/mlx4/en_rx.c
index 9b18d85..af26983 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_rx.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_rx.c
@@ -44,7 +44,7 @@
static int mlx4_en_alloc_frag(struct mlx4_en_priv *priv,
struct mlx4_en_rx_desc *rx_desc,
- struct skb_frag_struct *skb_frags,
+ struct page_frag *skb_frags,
struct mlx4_en_rx_alloc *ring_alloc,
int i)
{
@@ -61,7 +61,7 @@ static int mlx4_en_alloc_frag(struct mlx4_en_priv *priv,
return -ENOMEM;
skb_frags[i].page = page_alloc->page;
- skb_frags[i].page_offset = page_alloc->offset;
+ skb_frags[i].offset = page_alloc->offset;
page_alloc->page = page;
page_alloc->offset = frag_info->frag_align;
} else {
@@ -69,11 +69,11 @@ static int mlx4_en_alloc_frag(struct mlx4_en_priv *priv,
get_page(page);
skb_frags[i].page = page;
- skb_frags[i].page_offset = page_alloc->offset;
+ skb_frags[i].offset = page_alloc->offset;
page_alloc->offset += frag_info->frag_stride;
}
dma = pci_map_single(mdev->pdev, page_address(skb_frags[i].page) +
- skb_frags[i].page_offset, frag_info->frag_size,
+ skb_frags[i].offset, frag_info->frag_size,
PCI_DMA_FROMDEVICE);
rx_desc->data[i].addr = cpu_to_be64(dma);
return 0;
@@ -157,8 +157,8 @@ static int mlx4_en_prepare_rx_desc(struct mlx4_en_priv *priv,
struct mlx4_en_rx_ring *ring, int index)
{
struct mlx4_en_rx_desc *rx_desc = ring->buf + (index * ring->stride);
- struct skb_frag_struct *skb_frags = ring->rx_info +
- (index << priv->log_rx_info);
+ struct page_frag *skb_frags = ring->rx_info +
+ (index << priv->log_rx_info);
int i;
for (i = 0; i < priv->num_frags; i++)
@@ -183,7 +183,7 @@ static void mlx4_en_free_rx_desc(struct mlx4_en_priv *priv,
int index)
{
struct mlx4_en_dev *mdev = priv->mdev;
- struct skb_frag_struct *skb_frags;
+ struct page_frag *skb_frags;
struct mlx4_en_rx_desc *rx_desc = ring->buf + (index << ring->log_stride);
dma_addr_t dma;
int nr;
@@ -194,7 +194,7 @@ static void mlx4_en_free_rx_desc(struct mlx4_en_priv *priv,
dma = be64_to_cpu(rx_desc->data[nr].addr);
en_dbg(DRV, priv, "Unmapping buffer at dma:0x%llx\n", (u64) dma);
- pci_unmap_single(mdev->pdev, dma, skb_frag_size(&skb_frags[nr]),
+ pci_unmap_single(mdev->pdev, dma, skb_frags[nr].size,
PCI_DMA_FROMDEVICE);
put_page(skb_frags[nr].page);
}
@@ -403,7 +403,7 @@ void mlx4_en_deactivate_rx_ring(struct mlx4_en_priv *priv,
/* Unmap a completed descriptor and free unused pages */
static int mlx4_en_complete_rx_desc(struct mlx4_en_priv *priv,
struct mlx4_en_rx_desc *rx_desc,
- struct skb_frag_struct *skb_frags,
+ struct page_frag *skb_frags,
struct skb_frag_struct *skb_frags_rx,
struct mlx4_en_rx_alloc *page_alloc,
int length)
@@ -420,9 +420,9 @@ static int mlx4_en_complete_rx_desc(struct mlx4_en_priv *priv,
break;
/* Save page reference in skb */
- skb_frags_rx[nr].page = skb_frags[nr].page;
- skb_frag_size_set(&skb_frags_rx[nr], skb_frag_size(&skb_frags[nr]));
- skb_frags_rx[nr].page_offset = skb_frags[nr].page_offset;
+ __skb_frag_set_page(&skb_frags_rx[nr], skb_frags[nr].page);
+ skb_frag_size_set(&skb_frags_rx[nr], skb_frags[nr].size);
+ skb_frags_rx[nr].page_offset = skb_frags[nr].offset;
dma = be64_to_cpu(rx_desc->data[nr].addr);
/* Allocate a replacement page */
@@ -444,7 +444,7 @@ fail:
* the descriptor) of this packet; remaining fragments are reused... */
while (nr > 0) {
nr--;
- put_page(skb_frags_rx[nr].page);
+ __skb_frag_unref(&skb_frags_rx[nr]);
}
return 0;
}
@@ -452,7 +452,7 @@ fail:
static struct sk_buff *mlx4_en_rx_skb(struct mlx4_en_priv *priv,
struct mlx4_en_rx_desc *rx_desc,
- struct skb_frag_struct *skb_frags,
+ struct page_frag *skb_frags,
struct mlx4_en_rx_alloc *page_alloc,
unsigned int length)
{
@@ -474,7 +474,7 @@ static struct sk_buff *mlx4_en_rx_skb(struct mlx4_en_priv *priv,
/* Get pointer to first fragment so we could copy the headers into the
* (linear part of the) skb */
- va = page_address(skb_frags[0].page) + skb_frags[0].page_offset;
+ va = page_address(skb_frags[0].page) + skb_frags[0].offset;
if (length <= SMALL_PACKET_SIZE) {
/* We are copying all relevant data to the skb - temporarily
@@ -533,7 +533,7 @@ int mlx4_en_process_rx_cq(struct net_device *dev, struct mlx4_en_cq *cq, int bud
struct mlx4_en_priv *priv = netdev_priv(dev);
struct mlx4_cqe *cqe;
struct mlx4_en_rx_ring *ring = &priv->rx_ring[cq->ring];
- struct skb_frag_struct *skb_frags;
+ struct page_frag *skb_frags;
struct mlx4_en_rx_desc *rx_desc;
struct sk_buff *skb;
int index;
diff --git a/drivers/net/ethernet/mellanox/mlx4/en_tx.c b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
index 75dda26..75338eb 100644
--- a/drivers/net/ethernet/mellanox/mlx4/en_tx.c
+++ b/drivers/net/ethernet/mellanox/mlx4/en_tx.c
@@ -460,26 +460,13 @@ static inline void mlx4_en_xmit_poll(struct mlx4_en_priv *priv, int tx_ind)
}
}
-static void *get_frag_ptr(struct sk_buff *skb)
-{
- struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[0];
- struct page *page = frag->page;
- void *ptr;
-
- ptr = page_address(page);
- if (unlikely(!ptr))
- return NULL;
-
- return ptr + frag->page_offset;
-}
-
static int is_inline(struct sk_buff *skb, void **pfrag)
{
void *ptr;
if (inline_thold && !skb_is_gso(skb) && skb->len <= inline_thold) {
if (skb_shinfo(skb)->nr_frags == 1) {
- ptr = get_frag_ptr(skb);
+ ptr = skb_frag_address_safe(&skb_shinfo(skb)->frags[0]);
if (unlikely(!ptr))
return 0;
@@ -756,8 +743,9 @@ netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev)
/* Map fragments */
for (i = skb_shinfo(skb)->nr_frags - 1; i >= 0; i--) {
frag = &skb_shinfo(skb)->frags[i];
- dma = pci_map_page(mdev->dev->pdev, frag->page, frag->page_offset,
- skb_frag_size(frag), PCI_DMA_TODEVICE);
+ dma = skb_frag_dma_map(&mdev->dev->pdev->dev, frag,
+ 0, skb_frag_size(frag),
+ DMA_TO_DEVICE);
data->addr = cpu_to_be64(dma);
data->lkey = cpu_to_be32(mdev->mr.key);
wmb();
--
1.7.2.5
^ permalink raw reply related
* [PATCH 4/6] cxgb4vf: convert to SKB paged frag API.
From: Ian Campbell @ 2011-10-20 9:01 UTC (permalink / raw)
To: netdev@vger.kernel.org; +Cc: Ian Campbell, Casey Leedom, netdev
In-Reply-To: <1319101275.3385.129.camel@zakaz.uk.xensource.com>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Casey Leedom <leedom@chelsio.com>
Cc: netdev@vger.kernel.org
---
drivers/net/ethernet/chelsio/cxgb4vf/adapter.h | 2 +-
drivers/net/ethernet/chelsio/cxgb4vf/sge.c | 92 ++++++++++-------------
2 files changed, 41 insertions(+), 53 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4vf/adapter.h b/drivers/net/ethernet/chelsio/cxgb4vf/adapter.h
index 594334d..611396c 100644
--- a/drivers/net/ethernet/chelsio/cxgb4vf/adapter.h
+++ b/drivers/net/ethernet/chelsio/cxgb4vf/adapter.h
@@ -144,7 +144,7 @@ struct sge_fl {
* An ingress packet gather list.
*/
struct pkt_gl {
- skb_frag_t frags[MAX_SKB_FRAGS];
+ struct page_frag frags[MAX_SKB_FRAGS];
void *va; /* virtual address of first byte */
unsigned int nfrags; /* # of fragments */
unsigned int tot_len; /* total length of fragments */
diff --git a/drivers/net/ethernet/chelsio/cxgb4vf/sge.c b/drivers/net/ethernet/chelsio/cxgb4vf/sge.c
index c2d456d..8d5d55a 100644
--- a/drivers/net/ethernet/chelsio/cxgb4vf/sge.c
+++ b/drivers/net/ethernet/chelsio/cxgb4vf/sge.c
@@ -296,8 +296,8 @@ static int map_skb(struct device *dev, const struct sk_buff *skb,
si = skb_shinfo(skb);
end = &si->frags[si->nr_frags];
for (fp = si->frags; fp < end; fp++) {
- *++addr = dma_map_page(dev, fp->page, fp->page_offset,
- skb_frag_size(fp), DMA_TO_DEVICE);
+ *++addr = skb_frag_dma_map(dev, fp, 0, skb_frag_size(fp),
+ DMA_TO_DEVICE);
if (dma_mapping_error(dev, *addr))
goto unwind;
}
@@ -1357,6 +1357,35 @@ out_free:
}
/**
+ * copy_frags - copy fragments from gather list into skb_shared_info
+ * @skb: destination skb
+ * @gl: source internal packet gather list
+ * @offset: packet start offset in first page
+ *
+ * Copy an internal packet gather list into a Linux skb_shared_info
+ * structure.
+ */
+static inline void copy_frags(struct sk_buff *skb,
+ const struct pkt_gl *gl,
+ unsigned int offset)
+{
+ int i;
+
+ /* usually there's just one frag */
+ __skb_fill_page_desc(skb, 0, gl->frags[0].page,
+ gl->frags[0].offset + offset,
+ gl->frags[0].size - offset);
+ skb_shinfo(skb)->nr_frags = gl->nfrags;
+ for (i = 1; i < gl->nfrags; i++)
+ __skb_fill_page_desc(skb, i, gl->frags[i].page,
+ gl->frags[i].offset,
+ gl->frags[i].size);
+
+ /* get a reference to the last page, we don't own it */
+ get_page(gl->frags[gl->nfrags - 1].page);
+}
+
+/**
* t4vf_pktgl_to_skb - build an sk_buff from a packet gather list
* @gl: the gather list
* @skb_len: size of sk_buff main body if it carries fragments
@@ -1369,7 +1398,6 @@ struct sk_buff *t4vf_pktgl_to_skb(const struct pkt_gl *gl,
unsigned int skb_len, unsigned int pull_len)
{
struct sk_buff *skb;
- struct skb_shared_info *ssi;
/*
* If the ingress packet is small enough, allocate an skb large enough
@@ -1396,21 +1424,10 @@ struct sk_buff *t4vf_pktgl_to_skb(const struct pkt_gl *gl,
__skb_put(skb, pull_len);
skb_copy_to_linear_data(skb, gl->va, pull_len);
- ssi = skb_shinfo(skb);
- ssi->frags[0].page = gl->frags[0].page;
- ssi->frags[0].page_offset = gl->frags[0].page_offset + pull_len;
- skb_frag_size_set(&ssi->frags[0], skb_frag_size(&gl->frags[0]) - pull_len);
- if (gl->nfrags > 1)
- memcpy(&ssi->frags[1], &gl->frags[1],
- (gl->nfrags-1) * sizeof(skb_frag_t));
- ssi->nr_frags = gl->nfrags;
-
+ copy_frags(skb, gl, pull_len);
skb->len = gl->tot_len;
skb->data_len = skb->len - pull_len;
skb->truesize += skb->data_len;
-
- /* Get a reference for the last page, we don't own it */
- get_page(gl->frags[gl->nfrags - 1].page);
}
out:
@@ -1434,35 +1451,6 @@ void t4vf_pktgl_free(const struct pkt_gl *gl)
}
/**
- * copy_frags - copy fragments from gather list into skb_shared_info
- * @si: destination skb shared info structure
- * @gl: source internal packet gather list
- * @offset: packet start offset in first page
- *
- * Copy an internal packet gather list into a Linux skb_shared_info
- * structure.
- */
-static inline void copy_frags(struct skb_shared_info *si,
- const struct pkt_gl *gl,
- unsigned int offset)
-{
- unsigned int n;
-
- /* usually there's just one frag */
- si->frags[0].page = gl->frags[0].page;
- si->frags[0].page_offset = gl->frags[0].page_offset + offset;
- skb_frag_size_set(&si->frags[0], skb_frag_size(&gl->frags[0]) - offset);
- si->nr_frags = gl->nfrags;
-
- n = gl->nfrags - 1;
- if (n)
- memcpy(&si->frags[1], &gl->frags[1], n * sizeof(skb_frag_t));
-
- /* get a reference to the last page, we don't own it */
- get_page(gl->frags[n].page);
-}
-
-/**
* do_gro - perform Generic Receive Offload ingress packet processing
* @rxq: ingress RX Ethernet Queue
* @gl: gather list for ingress packet
@@ -1484,7 +1472,7 @@ static void do_gro(struct sge_eth_rxq *rxq, const struct pkt_gl *gl,
return;
}
- copy_frags(skb_shinfo(skb), gl, PKTSHIFT);
+ copy_frags(skb, gl, PKTSHIFT);
skb->len = gl->tot_len - PKTSHIFT;
skb->data_len = skb->len;
skb->truesize += skb->data_len;
@@ -1667,7 +1655,7 @@ int process_responses(struct sge_rspq *rspq, int budget)
rmb();
rsp_type = RSPD_TYPE(rc->type_gen);
if (likely(rsp_type == RSP_TYPE_FLBUF)) {
- skb_frag_t *fp;
+ struct page_frag *fp;
struct pkt_gl gl;
const struct rx_sw_desc *sdesc;
u32 bufsz, frag;
@@ -1701,9 +1689,9 @@ int process_responses(struct sge_rspq *rspq, int budget)
sdesc = &rxq->fl.sdesc[rxq->fl.cidx];
bufsz = get_buf_size(sdesc);
fp->page = sdesc->page;
- fp->page_offset = rspq->offset;
- skb_frag_size_set(fp, min(bufsz, len));
- len -= skb_frag_size(fp);
+ fp->offset = rspq->offset;
+ fp->size = min(bufsz, len);
+ len -= fp->size;
if (!len)
break;
unmap_rx_buf(rspq->adapter, &rxq->fl);
@@ -1717,9 +1705,9 @@ int process_responses(struct sge_rspq *rspq, int budget)
*/
dma_sync_single_for_cpu(rspq->adapter->pdev_dev,
get_buf_addr(sdesc),
- skb_frag_size(fp), DMA_FROM_DEVICE);
+ fp->size, DMA_FROM_DEVICE);
gl.va = (page_address(gl.frags[0].page) +
- gl.frags[0].page_offset);
+ gl.frags[0].offset);
prefetch(gl.va);
/*
@@ -1728,7 +1716,7 @@ int process_responses(struct sge_rspq *rspq, int budget)
*/
ret = rspq->handler(rspq, rspq->cur_desc, &gl);
if (likely(ret == 0))
- rspq->offset += ALIGN(skb_frag_size(fp), FL_ALIGN);
+ rspq->offset += ALIGN(fp->size, FL_ALIGN);
else
restore_rx_bufs(&gl, &rxq->fl, frag);
} else if (likely(rsp_type == RSP_TYPE_CPL)) {
--
1.7.2.5
^ permalink raw reply related
* [PATCH 6/6] net: add opaque struct around skb frag page
From: Ian Campbell @ 2011-10-20 9:01 UTC (permalink / raw)
To: netdev@vger.kernel.org; +Cc: Ian Campbell
In-Reply-To: <1319101275.3385.129.camel@zakaz.uk.xensource.com>
I've split this bit out of the skb frag destructor patch since it helps enforce
the use of the fragment API.
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
---
include/linux/skbuff.h | 10 ++++++----
net/core/skbuff.c | 6 +++---
2 files changed, 9 insertions(+), 7 deletions(-)
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 1ebf1ea..aec73c1 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -140,7 +140,9 @@ struct sk_buff;
typedef struct skb_frag_struct skb_frag_t;
struct skb_frag_struct {
- struct page *page;
+ struct {
+ struct page *p;
+ } page;
#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
__u32 page_offset;
__u32 size;
@@ -1175,7 +1177,7 @@ static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
{
skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
- frag->page = page;
+ frag->page.p = page;
frag->page_offset = off;
skb_frag_size_set(frag, size);
}
@@ -1699,7 +1701,7 @@ static inline void netdev_free_page(struct net_device *dev, struct page *page)
*/
static inline struct page *skb_frag_page(const skb_frag_t *frag)
{
- return frag->page;
+ return frag->page.p;
}
/**
@@ -1785,7 +1787,7 @@ static inline void *skb_frag_address_safe(const skb_frag_t *frag)
*/
static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
{
- frag->page = page;
+ frag->page.p = page;
}
/**
diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index e271040..ca4db40 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -668,14 +668,14 @@ int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask)
/* skb frags release userspace buffers */
for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
- put_page(skb_shinfo(skb)->frags[i].page);
+ skb_frag_unref(skb, i);
uarg->callback(uarg);
/* skb frags point to kernel buffers */
for (i = skb_shinfo(skb)->nr_frags; i > 0; i--) {
- skb_shinfo(skb)->frags[i - 1].page_offset = 0;
- skb_shinfo(skb)->frags[i - 1].page = head;
+ __skb_fill_page_desc(skb, i-1, head, 0,
+ skb_shinfo(skb)->frags[i - 1].size);
head = (struct page *)head->private;
}
--
1.7.2.5
^ permalink raw reply related
* [PATCH 3/6] cxgb4: convert to SKB paged frag API.
From: Ian Campbell @ 2011-10-20 9:01 UTC (permalink / raw)
To: netdev@vger.kernel.org; +Cc: Ian Campbell, Dimitris Michailidis, netdev
In-Reply-To: <1319101275.3385.129.camel@zakaz.uk.xensource.com>
Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
Cc: Dimitris Michailidis <dm@chelsio.com>
Cc: netdev@vger.kernel.org
---
drivers/net/ethernet/chelsio/cxgb4/cxgb4.h | 2 +-
drivers/net/ethernet/chelsio/cxgb4/sge.c | 45 ++++++++++++++-------------
2 files changed, 24 insertions(+), 23 deletions(-)
diff --git a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
index 223a7f7..0fe1885 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
+++ b/drivers/net/ethernet/chelsio/cxgb4/cxgb4.h
@@ -326,7 +326,7 @@ struct sge_fl { /* SGE free-buffer queue state */
/* A packet gather list */
struct pkt_gl {
- skb_frag_t frags[MAX_SKB_FRAGS];
+ struct page_frag frags[MAX_SKB_FRAGS];
void *va; /* virtual address of first byte */
unsigned int nfrags; /* # of fragments */
unsigned int tot_len; /* total length of fragments */
diff --git a/drivers/net/ethernet/chelsio/cxgb4/sge.c b/drivers/net/ethernet/chelsio/cxgb4/sge.c
index 14f31d3..ddc1698 100644
--- a/drivers/net/ethernet/chelsio/cxgb4/sge.c
+++ b/drivers/net/ethernet/chelsio/cxgb4/sge.c
@@ -215,8 +215,8 @@ static int map_skb(struct device *dev, const struct sk_buff *skb,
end = &si->frags[si->nr_frags];
for (fp = si->frags; fp < end; fp++) {
- *++addr = dma_map_page(dev, fp->page, fp->page_offset,
- skb_frag_size(fp), DMA_TO_DEVICE);
+ *++addr = skb_frag_dma_map(dev, fp, 0, skb_frag_size(fp),
+ DMA_TO_DEVICE);
if (dma_mapping_error(dev, *addr))
goto unwind;
}
@@ -1409,22 +1409,23 @@ int cxgb4_ofld_send(struct net_device *dev, struct sk_buff *skb)
}
EXPORT_SYMBOL(cxgb4_ofld_send);
-static inline void copy_frags(struct skb_shared_info *ssi,
+static inline void copy_frags(struct sk_buff *skb,
const struct pkt_gl *gl, unsigned int offset)
{
- unsigned int n;
+ int i;
/* usually there's just one frag */
- ssi->frags[0].page = gl->frags[0].page;
- ssi->frags[0].page_offset = gl->frags[0].page_offset + offset;
- skb_frag_size_set(&ssi->frags[0], skb_frag_size(&gl->frags[0]) - offset);
- ssi->nr_frags = gl->nfrags;
- n = gl->nfrags - 1;
- if (n)
- memcpy(&ssi->frags[1], &gl->frags[1], n * sizeof(skb_frag_t));
+ __skb_fill_page_desc(skb, 0, gl->frags[0].page,
+ gl->frags[0].offset + offset,
+ gl->frags[0].size - offset);
+ skb_shinfo(skb)->nr_frags = gl->nfrags;
+ for (i = 1; i < gl->nfrags; i++)
+ __skb_fill_page_desc(skb, i, gl->frags[i].page,
+ gl->frags[i].offset,
+ gl->frags[i].size);
/* get a reference to the last page, we don't own it */
- get_page(gl->frags[n].page);
+ get_page(gl->frags[gl->nfrags - 1].page);
}
/**
@@ -1459,7 +1460,7 @@ struct sk_buff *cxgb4_pktgl_to_skb(const struct pkt_gl *gl,
__skb_put(skb, pull_len);
skb_copy_to_linear_data(skb, gl->va, pull_len);
- copy_frags(skb_shinfo(skb), gl, pull_len);
+ copy_frags(skb, gl, pull_len);
skb->len = gl->tot_len;
skb->data_len = skb->len - pull_len;
skb->truesize += skb->data_len;
@@ -1478,7 +1479,7 @@ EXPORT_SYMBOL(cxgb4_pktgl_to_skb);
static void t4_pktgl_free(const struct pkt_gl *gl)
{
int n;
- const skb_frag_t *p;
+ const struct page_frag *p;
for (p = gl->frags, n = gl->nfrags - 1; n--; p++)
put_page(p->page);
@@ -1522,7 +1523,7 @@ static void do_gro(struct sge_eth_rxq *rxq, const struct pkt_gl *gl,
return;
}
- copy_frags(skb_shinfo(skb), gl, RX_PKT_PAD);
+ copy_frags(skb, gl, RX_PKT_PAD);
skb->len = gl->tot_len - RX_PKT_PAD;
skb->data_len = skb->len;
skb->truesize += skb->data_len;
@@ -1698,7 +1699,7 @@ static int process_responses(struct sge_rspq *q, int budget)
rmb();
rsp_type = RSPD_TYPE(rc->type_gen);
if (likely(rsp_type == RSP_TYPE_FLBUF)) {
- skb_frag_t *fp;
+ struct page_frag *fp;
struct pkt_gl si;
const struct rx_sw_desc *rsd;
u32 len = ntohl(rc->pldbuflen_qid), bufsz, frags;
@@ -1717,9 +1718,9 @@ static int process_responses(struct sge_rspq *q, int budget)
rsd = &rxq->fl.sdesc[rxq->fl.cidx];
bufsz = get_buf_size(rsd);
fp->page = rsd->page;
- fp->page_offset = q->offset;
- skb_frag_size_set(fp, min(bufsz, len));
- len -= skb_frag_size(fp);
+ fp->offset = q->offset;
+ fp->size = min(bufsz, len);
+ len -= fp->size;
if (!len)
break;
unmap_rx_buf(q->adap, &rxq->fl);
@@ -1731,16 +1732,16 @@ static int process_responses(struct sge_rspq *q, int budget)
*/
dma_sync_single_for_cpu(q->adap->pdev_dev,
get_buf_addr(rsd),
- skb_frag_size(fp), DMA_FROM_DEVICE);
+ fp->size, DMA_FROM_DEVICE);
si.va = page_address(si.frags[0].page) +
- si.frags[0].page_offset;
+ si.frags[0].offset;
prefetch(si.va);
si.nfrags = frags + 1;
ret = q->handler(q, q->cur_desc, &si);
if (likely(ret == 0))
- q->offset += ALIGN(skb_frag_size(fp), FL_ALIGN);
+ q->offset += ALIGN(fp->size, FL_ALIGN);
else
restore_rx_bufs(&si, &rxq->fl, frags);
} else if (likely(rsp_type == RSP_TYPE_CPL)) {
--
1.7.2.5
^ permalink raw reply related
* [PATCH net-next] virtio_net: fix truesize underestimation
From: Eric Dumazet @ 2011-10-20 9:14 UTC (permalink / raw)
To: David Miller
Cc: netdev, Rusty Russell, Michael S. Tsirkin, virtualization,
Krishna Kumar
We must account in skb->truesize, the size of the fragments, not the
used part of them.
Doing this work is important to avoid unexpected OOM situations.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
CC: Rusty Russell <rusty@rustcorp.com.au>
CC: "Michael S. Tsirkin" <mst@redhat.com>
CC: virtualization@lists.linux-foundation.org
CC: Krishna Kumar <krkumar2@in.ibm.com>
---
drivers/net/virtio_net.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/net/virtio_net.c b/drivers/net/virtio_net.c
index abbf34f..765ab9a 100644
--- a/drivers/net/virtio_net.c
+++ b/drivers/net/virtio_net.c
@@ -150,6 +150,7 @@ static void set_skb_frag(struct sk_buff *skb, struct page *page,
skb->data_len += size;
skb->len += size;
+ skb->truesize += PAGE_SIZE;
skb_shinfo(skb)->nr_frags++;
*len -= size;
}
@@ -287,7 +288,6 @@ static void receive_buf(struct net_device *dev, void *buf, unsigned int len)
}
hdr = skb_vnet_hdr(skb);
- skb->truesize += skb->data_len;
u64_stats_update_begin(&stats->syncp);
stats->rx_bytes += skb->len;
^ permalink raw reply related
* RE: [patch net-next]alx: Atheros AR8131/AR8151/AR8152/AR8161 Ethernet driver
From: Ren, Cloud @ 2011-10-20 9:23 UTC (permalink / raw)
To: David Miller
Cc: Rodriguez, Luis, netdev@vger.kernel.org,
linux-kernel@vger.kernel.org
In-Reply-To: <20111020.044541.970282389722164761.davem@davemloft.net>
>From: <cloud.ren@Atheros.com>
>Date: Thu, 20 Oct 2011 14:46:24 +0800
>
>> +#define __far
>
>So much unused crap left in these header files, get rid of this stuff.
>
>+#define ALX_HW_WARN(_fmt, _args...) \
>+ ALX_HW_PRINTA(WARNING, _fmt, ## _args)
>+
>+#define ALX_HW_INFO(_fmt, _args...) \
>+ ALX_HW_PRINTA(INFO, _fmt, ## _args)
>+
>+#define ALX_HW_DBG(_fmt, _args...) \
>+ ALX_HW_PRINTA(DEBUG, _fmt, ## _args)
>+
>
>We told you to get rid of your customized debug logging interfaces, yet all of
>this stuff is still there.
>
>+/* delay function */
>+#define US_DELAY(_hw, _n) __US_DELAY(_n)
>+#define MS_DELAY(_hw, _n) __MS_DELAY(_n)
>+#define __US_DELAY(_n) udelay(_n)
>+#define __MS_DELAY(_n) mdelay(_n)
>
>Useless wrappers for standard kernel interfaces, kill this.
>
>+#define DEBUG_INFO(_a, _b)
>+#define DEBUG_INFOS(_a, _b)
>
>Again we told you to get rid of this stuff.
>
>I suspect it's going to take may rounds of feedback before this driver is
>anywhere near ready for inclusion.
>
>Please just submit it to staging and let it cook there for a couple weeks in the
>interests of our sanity.
As you saw, should I do the two following steps?
1. I firstly try to submit code to linux-staging.git.
2. After the driver have been accepted by linux-staging.git, I submit to net-next.git again.
^ permalink raw reply
* Re: [PATCH 0/6] skb fragment API: convert network drivers (part V, take 2)
From: David Miller @ 2011-10-20 9:23 UTC (permalink / raw)
To: Ian.Campbell; +Cc: netdev, linux-scsi, linux-mm
In-Reply-To: <1319101275.3385.129.camel@zakaz.uk.xensource.com>
From: Ian Campbell <Ian.Campbell@citrix.com>
Date: Thu, 20 Oct 2011 10:01:15 +0100
> The following series is the second attempt to convert a fifth (and
> hopefully final) batch of network drivers to the SKB pages fragment API
> introduced in 131ea6675c76.
Applied, thanks Ian.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Fight unfair telecom internet charges in Canada: sign http://stopthemeter.ca/
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>
^ permalink raw reply
* Re: [PATCH net-next] bnx2x: fix skb truesize underestimation
From: David Miller @ 2011-10-20 9:23 UTC (permalink / raw)
To: eric.dumazet; +Cc: netdev, eilong
In-Reply-To: <1319101223.3781.7.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC>
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Thu, 20 Oct 2011 11:00:23 +0200
> bnx2x allocates a full page per fragment.
>
> We must account in skb->truesize, the size of the fragment, not the used
> part of it.
>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
> CC: Eilon Greenstein <eilong@broadcom.com>
Applied.
^ permalink raw reply
* Re: [patch net-next]alx: Atheros AR8131/AR8151/AR8152/AR8161 Ethernet driver
From: David Miller @ 2011-10-20 9:25 UTC (permalink / raw)
To: cjren; +Cc: rodrigue, netdev, linux-kernel
In-Reply-To: <6349D7A510622448B1BA0967850A8438011CC21D@nasanexd02d.na.qualcomm.com>
From: "Ren, Cloud" <cjren@qca.qualcomm.com>
Date: Thu, 20 Oct 2011 09:23:07 +0000
> As you saw, should I do the two following steps?
> 1. I firstly try to submit code to linux-staging.git.
> 2. After the driver have been accepted by linux-staging.git, I submit to net-next.git again.
You submit and get it into staging so that it can sit there for some
time and get reviewed and improved by others.
One doesn't submit directly to net-next right after it gets into
staging, staging is a place where your driver lives while it still
smelly funky and needs more work.
^ permalink raw reply
* Re: [PATCH net-next] virtio_net: fix truesize underestimation
From: David Miller @ 2011-10-20 9:23 UTC (permalink / raw)
To: eric.dumazet; +Cc: netdev, rusty, mst, virtualization, krkumar2
In-Reply-To: <1319102086.3781.13.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC>
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Thu, 20 Oct 2011 11:14:46 +0200
> We must account in skb->truesize, the size of the fragments, not the
> used part of them.
>
> Doing this work is important to avoid unexpected OOM situations.
>
> Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Applied.
^ permalink raw reply
* Re: PROBLEM: System call 'sendmsg' of process ospfd (quagga) causes kernel oops
From: David Miller @ 2011-10-20 9:30 UTC (permalink / raw)
To: herbert; +Cc: eric.dumazet, evonlanthen, linux-kernel, netdev, timo.teras
In-Reply-To: <20111019080807.GA25099@gondor.apana.org.au>
From: Herbert Xu <herbert@gondor.hengli.com.au>
Date: Wed, 19 Oct 2011 10:08:07 +0200
> I think Eric's initial patch is probably the safest bet for rc10.
> We can then work on the proper fix for the next release.
There are two "initial patch", I wonder which one you mean.
There's his really first patch, which remoevs the lines in IP_GRE
which change dev->needed_headroom. I was under the impression we
were against doing that.
The other patch he posted duplicates the device attribute variable
caching in two functions.
My patch is just a tweak so that we only do this sequence in one
place, the new sock_alloc_send_skb_reserve() helper, instead of
in both the ipv4 and ipv6 RAW code.
So I'm a little confused what your suggestion for rc10 really
is :-)
^ permalink raw reply
* Re: PROBLEM: System call 'sendmsg' of process ospfd (quagga) causes kernel oops
From: Herbert Xu @ 2011-10-20 9:35 UTC (permalink / raw)
To: David Miller; +Cc: eric.dumazet, evonlanthen, linux-kernel, netdev, timo.teras
In-Reply-To: <20111020.053050.383972361986316046.davem@davemloft.net>
On Thu, Oct 20, 2011 at 05:30:50AM -0400, David Miller wrote:
>
> So I'm a little confused what your suggestion for rc10 really
> is :-)
I meant his first initial patch :)
While it is suboptimal in the sense that should the value of
needed_headroom increase we'll end up constantly reallocating
skbs, I believe that it is at least semantically correct.
In the time being I'll look more closely at all the users of
needed_headroom to see if there's anything we've missed.
Thanks,
--
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox