Instead of masking head and tail every time we increment them, just let them
wrap through UINT_MAX and mask them when subscripting. Add simple accessor
functions to do the subscripting properly to minimize the chances of messing
this up.
...
+static unsigned int macb_tx_ring_avail(struct macb
On Tue, Oct 30, 2012 at 4:12 AM, David Laight david.lai...@aculab.com wrote:
Instead of masking head and tail every time we increment them, just let
them
wrap through UINT_MAX and mask them when subscripting. Add simple accessor
functions to do the subscripting properly to minimize
return (TX_RING_SIZE - (bp-tx_head - bp-tx_tail) (TX_RING_SIZE -
1));
Is equivalent to:
return (bp-tx_tail - bp-tx_head) (TX_RING_SIZE - 1));
David
--
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to
On Tue, Oct 30, 2012 at 02:45:57PM -0400, Sasha Levin wrote:
+/* Use hash_32 when possible to allow for fast 32bit hashing in 64bit
kernels. */
+#define hash_min(val, bits)
\
+({
this_cpu_read
|-_this_cpu_generic_read
#define _this_cpu_generic_read(pcp) \
({ typeof(pcp) ret__; \
preempt_disable(); \
ret__ =
This patch adds proper handling of the buggy revision A2 of LXT973 phy, adding
precautions linked to ERRATA Item 4:
Revision A2 of LXT973 chip randomly returns the contents of the previous even
register when you read a odd register regularly
Does reading the PHY registers involve
Amazing how something simple gets lots of comments and versions :-)
...
+ * This has to be a macro since HASH_BITS() will not work on pointers since
+ * it calculates the size during preprocessing.
+ */
+#define hash_empty(hashtable)
\
And even then, if we would do:
for (i = 0; i HASH_SIZE(hashtable); i++)
if (!hlist_empty(hashtable[i]))
break;
return i = HASH_SIZE(hashtable);
What happens if the last entry of the table is non-empty ?
It still works, as 'i' is not
Moreover, if your thinking is that we do not need a static inline
function replicated at every caller, maybe we should introduce a
lib/hashtable.c that implements those 2 functions.
That was my thought...
Given their nature, I'd guess they aren't critical path.
Probably not worth adding an
On Sun, 04 Nov 2012 00:43:31 +0900, Akinobu Mita said:
This patchset introduces new functions into random32 library for
getting the requested number of pseudo-random bytes.
Before introducing these new functions into random32 library,
prandom32() and prandom32_seed() with prandom32
Looks the problem is worse than above, not only bitfields are affected, the
adjacent fields might be involved too, see:
http://lwn.net/Articles/478657/
Not mentioned in there is that even with x86/amd64 given
a struct with the following adjacent fields:
char a;
- if (pkt_dev-min_in6_daddr.s6_addr32[0] == 0
- pkt_dev-min_in6_daddr.s6_addr32[1] == 0
- pkt_dev-min_in6_daddr.s6_addr32[2] == 0
- pkt_dev-min_in6_daddr.s6_addr32[3] == 0) ;
- else {
+ if
ipv4 and ipv6 use different styles for these tests.
ipv4_is_foo(__be32)
ipv6_addr_foo(struct in6_addr *)
I presume there is a 'const' in there ...
Perhaps it'd be good to convert the ipv4 tests to the ipv6 style.
You don't want to force an IPv4 address (which might be in a register)
be
...
long is always the same or bigger then a pointer
(A pointer must always fit in a long)
...
Linux may make that assumption, but it doesn't have
to be true. 64bit windows still has 32bit long.
C99 inttypes.h defines [u]intptr_t to be an integral type
that is large enough to hold a pointer
Between two invocations of make_eligible, the system virtual time may
happen to grow enough that, in its binary representation, a bit with
higher order than 31 flips. This happens especially with
TSO/GSO. Before this fix, the mask used in make_eligible was computed
as
Allocate regular pages to use as backing for the RX ring and use the
DMA API to sync the caches. This should give a bit better performance
since it allows the CPU to do burst transfers from memory. It is also
a necessary step on the way to reduce the amount of copying done by
the driver.
On 12/03/2012 01:43 PM, David Laight :
Allocate regular pages to use as backing for the RX ring and use the
DMA API to sync the caches. This should give a bit better performance
since it allows the CPU to do burst transfers from memory. It is also
a necessary step on the way to reduce
If I understand well, you mean that the call to:
dma_sync_single_range_for_device(bp-pdev-dev, phys,
pg_offset, frag_len, DMA_FROM_DEVICE);
in the rx path after having copied the data to skb is not needed?
That is also the conclusion that I
Well, for the 10/100 MACB interface, I am stuck with 128 Bytes buffers!
So this use of pages seems sensible.
If you have dma coherent memory you can make the rx buffer space
be an array of short buffers referenced by adjacent ring entries
(possibly with the last one slightly short to allow for
+struct ax88179_rx_pkt_header {
+ u8 l4_csum_err:1,
+ l3_csum_err:1,
+ l4_type:3,
+ l3_type:2,
+ ce:1;
+
+ u8 vlan_ind:3,
+ rx_ok:1,
+ pri:3,
+ bmc:1;
+
+ u16 len:13,
+
On Thu, Feb 28, 2013 at 01:53:25PM -0800, Andy Lutomirski wrote:
O_DENYMAND - to switch on/off three flags above.
O_DENYMAND doesn't deny anything. Would a name like O_RESPECT_DENY be
better?
Possibly rename to O_CHECK_DENY ?
David
--
David Laight: da...@l8s.co.uk
+ n = kmalloc(sizeof *n, GFP_KERNEL | __GFP_NOWARN);
+ if (!n)
+ n = vmalloc(sizeof *n);
I'm slightly confused by this construct.
I thought kmalloc(... GFP_KERNEL) would sleep waiting for
memory (rather than return NULL).
I realise that (for multi-page sizes) that
I think this means that kmalloc() is likely to be faster
(if it doesn't have to sleep), but that vmalloc() is
allocating from a much larger resource.
This make me that that large allocations that are not
temporary should probably be allocated with vmalloc().
vmalloc has some issues
+ int gain_data[] = {0x0062, 0x0064, 0x006a, 0x106a, 0x106c,
+0x1074, 0x107c, 0x207c};
static const int ...
David
--
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to
I wouldn't go that far... ;-) Unfairness is not a show-stopper right?
IMHO, the warning/documentation should suffice for anybody wanting to
try out this locking scheme for other use-cases.
I presume that by 'fairness' you mean 'write preference'?
I'd not sure how difficult it would be, but
A proper solution would be to either return false if net.ipv6.bindv6only is
true and optval is false
(which would break downward compatibility because it wouldn't just be a
default and setsockopt might
return an error) or to introduce a new sysctl variable like
My solution to making 'break' work in the iterator is:
for (bkt = 0, node = NULL; bkt HASH_SIZE(name) node ==
NULL; bkt++)
hlist_for_each_entry(obj, node, name[bkt], member)
I'd take a look at the generated code.
Might come out a bit better if the condition is changed
- for (i = 0;
- old_priomap (i old_priomap-priomap_len);
- i++)
- new_priomap-priomap[i] = old_priomap-priomap[i];
+ if (old_priomap) {
+ old_len = old_priomap-priomap_len;
+
+ for (i = 0; i old_len; i++)
+
Remove useless kfree() and clean up code related to the removal.
...
diff --git a/drivers/isdn/gigaset/common.c b/drivers/isdn/gigaset/common.c
index aa41485..30a6b17 100644
--- a/drivers/isdn/gigaset/common.c
+++ b/drivers/isdn/gigaset/common.c
@@ -1123,7 +1123,6 @@ struct gigaset_driver
+ /* Read command completed register */
+ done_mask = ioread32(hcr_base + CC);
+
+ if (host_priv-quirks SATA_FSL_QUIRK_V2_ERRATA) {
+ if (unlikely(hstatus INT_ON_DATA_LENGTH_MISMATCH)) {
+ for (tag = 0; tag ATA_MAX_QUEUE; tag++) {
+
Yes hash_32 seems reasonable for the uid hash. With those long hash
chains I wouldn't like to be on a machine with 10,000 processes with
each with a different uid, and a processes calling setuid in the fast
path.
The uid hash that we are playing with is one that I sort of wish that
the
I would suggest to increase TCP_SYN_RETRIES from 5 to 6.
180 secs is eternity, but 31 secs is too small.
Wasn't the intention of the long delay to allow a system
acting as a router to reboot?
I suspect that is why it (and some other TCP timers)
are in minutes.
David
On Tue, 27 Nov 2012 18:02:29 +
David Woodhouse dw...@infradead.org wrote:
In solos-pci at least, the ops-close() function doesn't flush all
pending skbs for this vcc before returning. So can be a tasklet
somewhere which has loaded the address of the vcc-pop function from one
of
compiler manages to obtain exclusive access
to the source files - stopping the other systems from reading them.
David
--
David Laight: da...@l8s.co.uk
--
To unsubscribe from this list: send the line unsubscribe linux-kernel in
the body of a message to majord...@vger.kernel.org
More majordomo
Subject: Re: [PATCH 0/2] fs: supply inode uid/gid setting interface
On 2013/8/23 12:10, Greg KH wrote:
On Fri, Aug 23, 2013 at 10:48:36AM +0800, Rui Xiang wrote:
This patchset implements an accessor functions to set uid/gid
in inode struct. Just finish code clean up.
Why?
It can
+static void netif_free_tx_queues(struct net_device *dev)
+{
+ if (is_vmalloc_addr(dev-_tx))
+ vfree(dev-_tx);
+ else
+ kfree(dev-_tx);
+}
+
static int netif_alloc_netdev_queues(struct net_device *dev)
{
unsigned int count = dev-num_tx_queues;
@@
I think we should be able to do something better using the C
preprocessor, this is exactly the sort of thing it's good at.
What I mean is something like we do with arch/powerpc/include/asm/systbl.h,
where we define the list of syscalls once, and then include it in
multiple places, using
So it should be declared dma_addr_t then,
+ addr = dma_map_single(ndev-dev, (void
*)rx_buff-skb-data,
+ buflen, DMA_FROM_DEVICE);
+ if (dma_mapping_error(ndev-dev, addr)) {
+ if (net_ratelimit())
On Wed, Sep 11, 2013 at 05:04:17PM -0700, Joe Perches wrote:
On Thu, 2013-09-12 at 08:40 +0900, Tetsuo Handa wrote:
Joe Perches wrote:
- seq_printf(m, %s%d%n, con-name, con-index, len);
+ len = seq_printf(m, %s%d, con-name, con-index);
Isn't len always 0 or -1 ?
I just did a diff of registers in exynos 4210 and 4212 PHY drivers [1]
and couldn't find that big a difference in register layout. Of course
there are a few changes in HSIC bit fields and PHYFSEL but that's only
minimal and could well be handled in a single driver.
[1] -
It seems to me that a more useful interface would take a minimum and
maximum number of vectors from the driver. This wouldn't allow the
driver to specify that it could only accept, say, any even number within
a certain range, but you could still leave the current functions
available for
Implement instr_is_load_store_2_06() to detect whether a given instruction
is one of the fixed-point or floating-point load/store instructions in the
POWER Instruction Set Architecture v2.06.
...
+int instr_is_load_store_2_06(const unsigned int *instr)
+{
+ unsigned int op, upper, lower;
On Tue, 2013-10-15 at 14:29 +, Kelleter, Günther wrote:
Addresses are BCD encoded, not ASCII. x25_addr_ntoa got it right.
[]
Wrong length calculation leads to rejection of CALL ACCEPT packets.
[]
diff --git a/net/x25/af_x25.c b/net/x25/af_x25.c
[]
@@ -98,7 +98,7 @@ int
Subject: [PATCH RFC v2 29/29] vmxnet3: Make use of pcim_enable_msix_range()
interface
...
diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c
b/drivers/net/vmxnet3/vmxnet3_drv.c
index d33802c..e552d2b 100644
--- a/drivers/net/vmxnet3/vmxnet3_drv.c
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@
From: Of Roger Quadros
With u-boot 2013.10, USB devices are sometimes not detected
on OMAP4 Panda. To make us independent of what bootloader does
with the USB Host module, we must RESET it to get it to a known
good state. This patch Soft RESETs the USB Host module.
...
+++
From: David Miller
According to Documentation/networking/driver.txt the ndo_start_xmit
method should not return NETDEV_TX_BUSY under normal circumstances.
Stop the transmit queue when tx_ring is full.
...
You should be instead preventing the transmit method from being invoked
when it
From: Roger Quadros [mailto:rog...@ti.com]
On 11/29/2013 03:17 PM, David Laight wrote:
...
+ timeout = jiffies + msecs_to_jiffies(100);
+ while (!(usbhs_read(omap-uhh_base, OMAP_UHH_SYSSTATUS)
+ OMAP_UHH_SYSSTATUS_RESETDONE)) {
+ cpu_relax();
You mean use
@@ -1637,7 +1637,7 @@ static int sync_thread_master(void *data)
continue;
}
while (ip_vs_send_sync_msg(tinfo-sock, sb-mesg) 0) {
- int ret = __wait_event_interruptible(*sk_sleep(sk),
So ideally there's be a comment here
I've done this in the past so that the code sleeps interruptibly
unless there is a signal pending - which would cause it to return
early.
/* Tell scheduler we are going to sleep... */
if (signal_pending(current))
/* We don't want waking immediately (again) */
Sure, I modified the code so that we only prefetched 2 cache lines ahead, but
only if the overall length of the input buffer is more than 2 cache lines.
Below are the results (all counts are the average of 100 iterations of the
csum operation, as previous tests were, I just omitted that
I'm not sure, whats the typical capacity for the branch predictors
ability to remember code paths?
...
For such simple single-target branches it goes near or over a thousand for
recent Intel and AMD microarchitectures. Thousands for really recent CPUs.
IIRC the x86 can also correctly
From: Eric Dumazet [mailto:eric.duma...@gmail.com]
On Tue, 2013-11-19 at 09:02 -0500, Mark Lord wrote:
On 13-11-19 05:04 AM, David Laight wrote:
From: Mark Lord
..
except the ax88179_178a driver still does not work in linux-3.12,
whereas it works fine in all earlier kernels.
I've
From: Eric Dumazet [mailto:eric.duma...@gmail.com]
On Tue, 2013-11-19 at 14:43 +, David Laight wrote:
It isn't directly a TSO problem.
There has always been a bug in the xhci driver for fragmented buffers.
TSO just means it is given a lot of fragmented buffers.
As well as user
In normal operation, IP-A sends INIT to IP-X, IP-X returns INIT_ACK to
IP-A. IP-A then sends HB to IP-X, IP-X then returns HB_ACK to IP-A. In
the meantime, IP-B sends HB to IP-Y and IPY returns HB_ACK.
In case of the path between IP-A and IP-X is broken, IP-B sends INIT
to IP-X, NODE-B
There are some network configurations that do cause problems.
Consider 4 systems with 3 LAN segments:
A) 10.10.10.1 on LAN X and 192.168.1.1 on LAN Y.
B) 10.10.10.2 on LAN X and 192.168.1.2 on LAN Y.
C) 10.10.10.3 on LAN X.
D) 10.10.10.4 on LAN X and 192.168.1.2 on LAN Z.
There are
The point is that address scoping should be used. When sending an
INIT from 10.10.10.1 to 10.10.10.4 you should not list 192.168.1.1,
since you are transmitting an address to a node which might or might
not be in the same scope.
You might have two machines that are connected via the public
the configured addresses could be:
System A) 10.0.0.1 on Lan X, 10.10.0.1 on Lan Y
System B) 10.0.0.2 on Lan X, 10.10.0.2 on Lan Y
System C) 10.0.0.3 on Lan X, 10.10.0.2 on Lan Z
Same problem will occur.
...
With that, Sys A talking to Sys C will get an abort
from Sys B when trying
From: Hayes Wang
For slow CPU, the frequent bulk transfer would cause poor throughput.
One solution is to increase the timeout of the aggregation. It let
the hw could complete the bulk transfer later and fill more packets
into the buffer. Besides, it could reduce the frequency of the bulk
From: Hayes Wang
...
I should have spotted this before.
/* USB_RX_EARLY_AGG */
-#define EARLY_AGG_SUPPER 0x0e832981
+#define EARLY_AGG_SUPER rx_buf_sz - 1522) / 4) 16) | \
+ (u32)(CONFIG_RTL8152_EARLY_AGG_TIMEOUT_SUPER = 0 ? 85 * 125 : \
+
From: David Herrmann
The name_assign_type attribute gives hints where the interface name of a
given net-device comes from. Three different values are currently defined:
NET_NAME_ENUM:
This is the default. The ifname is provided by the kernel with an
enumerated suffix. Names may be
From: beh...@converseincode.com
From: Mark Charlebois charl...@gmail.com
Replaced non-standard C use of Variable Length Arrays In Structs (VLAIS) in
xt_repldata.h with a C99 compliant flexible array member and then calculated
offsets to the other struct members. These other members aren't
From: Behan Webster
On 03/18/14 02:41, David Laight wrote:
From: beh...@converseincode.com
From: Mark Charlebois charl...@gmail.com
Replaced non-standard C use of Variable Length Arrays In Structs (VLAIS) in
xt_repldata.h with a C99 compliant flexible array member
From: Joe Perches
Reduce text a bit by using static const.
If you want to save a few bytes remove the pointers.
(and the fixed RAM text to get below 7 chars).
eg:
- const char *ramtypetext2[] = { SDR SDRAM, SDR SGRAM,
- DDR SDRAM, DDR SGRAM };
From: Joe Perches [
On Mon, 2014-02-24 at 10:26 +, David Laight wrote:
From: Joe Perches
Reduce text a bit by using static const.
If you want to save a few bytes remove the pointers.
(and the fixed RAM text to get below 7 chars).
Hi David.
eg:
- const char
From: Keith Packard
The Lenovo OneLink dock includes a USB ethernet adapter using the
AX88179 chip, but with a different USB ID. Add this new USB id to the
driver so that it will autodetect the adapter correctly.
Signed-off-by: Keith Packard kei...@keithp.com
Tested-by: Carl Worth
From: Hayes Wang
Support scatter gather and TSO.
Adjust the tx checksum function and set the max gso size to fix the
size of the tx aggregation buffer.
There is little point supporting TSO unless the usb host controller
supports arbitrary aligned scatter-gather.
All you do is require that a
From: hayeswang
David Laight [mailto:david.lai...@aculab.com]
Sent: Tuesday, March 04, 2014 8:12 PM
To: 'Hayes Wang'; net...@vger.kernel.org
Cc: nic_s...@realtek.com; linux-kernel@vger.kernel.org;
linux-...@vger.kernel.org
Subject: RE: [PATCH net-next 08/12] r8152: support TSO
From: Eric Dumazet
On Tue, 2014-03-04 at 14:35 +, David Laight wrote:
Does that mean you are splitting the 64k 'ethernet packet' from TCP
is software? I've looked at the ax88179 where the hardware can do it.
Is there really a gain doing segmentation here if you have to do
From: Bjorn Helgaas
With -Werror=array-bounds, gcc v4.7.x warns that in phy_find_valid(), the
settings[] array subscript is above array bounds, I think because idx is
a signed integer and if the caller supplied idx 0, we pass the guard but
still reference out of bounds.
Not rejecting the
From: Bjorn Helgaas
On Wed, Mar 5, 2014 at 2:10 AM, David Laight david.lai...@aculab.com wrote:
From: Bjorn Helgaas
With -Werror=array-bounds, gcc v4.7.x warns that in phy_find_valid(), the
settings[] array subscript is above array bounds, I think because idx is
a signed integer
From: Bjorn Helgaas
I'm stumped. phy_find_valid() is static and only called from one
place. The 'idx' argument is always the result of phy_find_setting(),
which should always return something between 0 and
ARRAY_SIZE(settings), so I don't see any way idx can be 0.
I stripped this down as
From: Sukadev Bhattiprolu
When checking whether a bit representing a register is set in
sample_regs, a 64-bit mask, use 64-bit value (1LL).
Signed-off-by: Sukadev Bhattiprolu suka...@linux.vnet.ibm.com
---
tools/perf/util/unwind.c |4 ++--
1 file changed, 2 insertions(+), 2
From: Mathias Nyman
This reverts commit 3804fad45411b48233b48003e33a78f290d227c8.
This commit, together with commit 247bf557273dd775505fb9240d2d152f4f20d304
xhci 1.0: Limit arbitrarily-aligned scatter gather. were
origially added to get xHCI 1.0 hosts and usb ethernet ax88179_178a devices
From: Mathias Nyman
This reverts commit 247bf557273dd775505fb9240d2d152f4f20d304.
You need to revert further.
Just don’t set hcd-self.no_sg_constraint ever - since it just
doesn't work.
That will stop the ax88179_178a driver sending fragmented packets.
With the check for aligned non-terminal
From: Alan Stern
On Fri, 7 Mar 2014, David Laight wrote:
From: Mathias Nyman
This reverts commit 3804fad45411b48233b48003e33a78f290d227c8.
This commit, together with commit 247bf557273dd775505fb9240d2d152f4f20d304
xhci 1.0: Limit arbitrarily-aligned scatter gather. were
From: Alan Stern
On Fri, 7 Mar 2014, David Laight wrote:
From: Mathias Nyman
This reverts commit 247bf557273dd775505fb9240d2d152f4f20d304.
You need to revert further.
Just don?t set hcd-self.no_sg_constraint ever - since it just
doesn't work.
No; it does work most of the time
From:
On Fri, 7 Mar 2014, David Laight wrote:
From: Alan Stern
On Fri, 7 Mar 2014, David Laight wrote:
From: Mathias Nyman
This reverts commit 3804fad45411b48233b48003e33a78f290d227c8.
This commit, together with commit
247bf557273dd775505fb9240d2d152f4f20d304
From: j...@joshtriplett.org
On Fri, Mar 07, 2014 at 01:02:44PM -0800, Joe Perches wrote:
On Fri, 2014-03-07 at 15:41 -0500, David Miller wrote:
From: Alistair Popple alist...@popple.id.au
Date: Thu, 6 Mar 2014 14:52:25 +1100
+ out_be32(dev-reg, in_be32(dev-reg) |
From: David Miller
From: Linus Torvalds torva...@linux-foundation.org
Date: Mon, 10 Mar 2014 19:02:18 -0700
I would generally suggest that people only use bool for function
return types, and absolutely nothing else. Seriously.
I think it makes sense for function arguments too.
'bool'
From: Zhouyi Zhou
not frequently changing components should share same cachelines
Signed-off-by: Zhouyi Zhou yizhouz...@ict.ac.cn
---
include/net/netns/conntrack.h | 12
1 file changed, 8 insertions(+), 4 deletions(-)
diff --git a/include/net/netns/conntrack.h
From: Shahed Shaikh
Adding netdev.
-Original Message-
From: Martin Kaiser,,, [mailto:mar...@reykholt.kaiser.cx] On Behalf Of
Martin Kaiser
Sent: Thursday, January 09, 2014 9:29 PM
To: Himanshu Madhani; Rajesh Borundia
Cc: linux-kernel; triv...@kernel.org
Subject: [PATCH
From: David Miller
...
On Thu, Jan 9, 2014 at 12:26 PM, Neal Cardwell ncardw...@google.com wrote:
On Thu, Jan 9, 2014 at 3:16 PM, Dan Williams dan.j.willi...@intel.com
wrote:
net_dma was the only external user so this can become local to tcp.c
again.
...
-void
From: walt
In the meantime, try this patch, which is something of a long shot.
No difference. But I notice the code enables the TRB quirk only if the
xhci_version is specifically 0x95. My debug messages claim that xHCI
doesn't need link TRB QUIRK so I'm wondering if adding my asmedia
From: Mathias Nyman
This is an attempt to re-work and solve the issues in xhci command
queue management that Sarah has descibed earlier:
Right now, the command management in the xHCI driver is rather ad-hock.
Different parts of the driver all submit commands, including interrupt
handling
From: walt
On 01/09/2014 03:50 PM, Sarah Sharp wrote:
On Tue, Jan 07, 2014 at 03:57:00PM -0800, walt wrote:
I've wondered if my xhci problems might be caused by hardware quirks, and
wondering why I seem to be the only one who has this problem.
Maybe I could take one for the team by
From: Mathias Nyman
...
The fact that you are having to allocate memory ion an ISR ought also to be
ringing alarm bells.
It did.
Other options are as you said to use a 'software command ring'. Either
just pre-allocate a full command ring (64 trbs * sizeof(struct
xhci_command)), roughly
From: Emil Goode
int rndis_rx_fixup(struct usbnet *dev, struct sk_buff *skb)
{
+ /* This check is no longer done by usbnet */
+ if (skb-len dev-net-hard_header_len)
+ return 0;
+
Wouldn't it be better to test against ETH_HLEN, since that is a constant
and
From: Of Emil Goode
This patch removes a generic hard_header_len check from the usbnet
module that is causing dropped packages under certain circumstances
for devices that send rx packets that cross urb boundaries.
One example is the AX88772B which occasionally send rx packets that
cross
+
+ /* Update the heartbeat timer immediately. */
+ if (!mod_timer(trans-hb_timer,
+ sctp_transport_timeout(trans)))
+ sctp_transport_hold(trans);
This is causes
From: Oliver Neukum
censored. Oh well. But how about merging it with FLAG_MULTI_PACKET?
I really don't want to add more flags. There is a point where enough
flags make absurd having a common code. We are closing in on that point.
Any sub-driver that supports multi-packet either has to use
From: Michal Simek
From: Peter Crosthwaite peter.crosthwa...@xilinx.com
The packet completion interrupts for TX and RX should be serviced before
the packets are consumed. This ensures against the degenerate case when a
new completion interrupt is raised after the handler has exited but
From: beh...@converseincode.com
From: Mark Charlebois charl...@gmail.com
Replaced non-standard C use of Variable Length Arrays In Structs (VLAIS) in
xt_repldata.h with a C99 compliant flexible array member and then calculated
offsets to the other struct members. These other members aren't
From: beh...@converseincode.com
Replaced the use of a Variable Length Array In Struct (VLAIS) with a C99
compliant equivalent. This is the original VLAIS struct.
struct {
struct aead_request req;
u8 priv[crypto_aead_reqsize(tfm)];
} aead_req;
This
From: Kevin Hao
Sent: 20 March 2014 11:48
To: Scott Wood
Cc: linuxppc-...@lists.ozlabs.org; Chenhui Zhao; jason@freescale.com;
linux-kernel@vger.kernel.org
Subject: Re: [PATCH 9/9] powerpc/pm: support deep sleep feature on T1040
On Tue, Mar 18, 2014 at 06:18:54PM -0500, Scott Wood
From: Scott Wood [mailto:scottw...@freescale.com]
On Thu, 2014-03-20 at 11:59 +, David Laight wrote:
I tried to work out what the 'twi, isync' instructions were for (in
in_le32()).
The best I could come up with was to ensure a synchronous bus-fault.
But bus faults are probably only
From: Eric Dumazet
On Thu, 2014-03-20 at 18:56 -0700, Andi Kleen wrote:
Eric Dumazet eric.duma...@gmail.com writes:
I saw csum_partial() consuming 1% of cpu cycles in a GRO workload, that
is insane...
Couldn't it just be the cache miss?
Or the fact that we mix 16 bit stores
EXPORT_SYMBOL and inline directives are contradictory to each other.
The patch fixes this inconsistency.
...
-inline const struct nf_nat_l4proto *
+const struct nf_nat_l4proto *
__nf_nat_l4proto_find(u8 family, u8 protonum)
{
return
Looks like the ptype_base[] should be per 'dev'?
Or just put entries where ptype-dev != null_or_dev on a per-interface
list and do two searches?
Yes, but then we would have two searches instead of one in fast path.
Usually it would be empty - so the search would be very quick!
I am a bit uneasy with this one, because an applicatio polling() on one
thousand file descriptors using select()/poll(), will call sk_poll_ll()
one thousand times.
Anything calling poll() on 1000 fds probably has performance
issues already! Which is why kevent schemes have been added.
At
I have a Linux router with a lot of interfaces (hundreds or
thousands of VLANs) and an application that creates AF_PACKET
socket per interface and bind()s sockets to interfaces.
...
I noticed that box has strange performance problems with
most of the CPU time spent in
1 - 100 of 2841 matches
Mail list logo
