I found the PPP subsystem to not work properly when connecting channels
with different speeds to the same bundle.
Problem Description:
As the "ppp_mp_explode" function fragments the sk_buff buffer evenly
among the PPP channels that are connected to a certain PPP unit to
make up a bundle, if we are transmitting using an upper layer protocol
that requires an Ack before sending the next packet (like TCP/IP for
example), we will have a bandwidth bottleneck on the slowest channel
of the bundle.
Let's clarify by an example. Let's consider a scenario where we have
two PPP links making up a bundle: a slow link (10KB/sec) and a fast
link (1000KB/sec) working at the best (full bandwidth). On the top we
have a TCP/IP stack sending a 1000 Bytes sk_buff buffer down to the
PPP subsystem. The "ppp_mp_explode" function will divide the buffer in
two fragments of 500B each (we are neglecting all the headers, crc,
flags etc?.). Before the TCP/IP stack sends out the next buffer, it
will have to wait for the ACK response from the remote peer, so it
will have to wait for both fragments to have been sent over the two
PPP links, received by the remote peer and reconstructed. The
resulting behaviour is that, rather than having a bundle working
@1010KB/sec (the sum of the channels bandwidths), we'll have a bundle
working @20KB/sec (the double of the slowest channels bandwidth).
Problem Solution:
The problem has been solved by redesigning the "ppp_mp_explode"
function in such a way to make it split the sk_buff buffer according
to the speeds of the underlying PPP channels (the speeds of the serial
interfaces respectively attached to the PPP channels). Referring to
the above example, the redesigned "ppp_mp_explode" function will now
divide the 1000 Bytes buffer into two fragments whose sizes are set
according to the speeds of the channels where they are going to be
sent on (e.g . 10 Byets on 10KB/sec channel and 990 Bytes on
1000KB/sec channel). The reworked function grants the same
performances of the original one in optimal working conditions (i.e. a
bundle made up of PPP links all working at the same speed), while
greatly improving performances on the bundles made up of channels
working at different speeds.
Signed-off-by: Gabriele Paoloni <gabriele.paoloni@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Remove some pointless conditionals before kfree_skb().
Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In each case, ap is checked not to be NULL a few lines before.
A simplified version of the semantic patch that makes this change is as
follows: (http://www.emn.fr/x-info/coccinelle/)
// <smpl>
@r exists@
local idexpression x;
expression E;
position p1,p2;
@@
if (x@p1 == NULL || ...) { ... when forall
return ...; }
... when != \(x=E\|x--\|x++\|--x\|++x\|x-=E\|x+=E\|x|=E\|x&=E\|&x\)
(
x@p2 == NULL
|
x@p2 != NULL
)
// another path to the test that is not through p1?
@s exists@
local idexpression r.x;
position r.p1,r.p2;
@@
... when != x@p1
(
x@p2 == NULL
|
x@p2 != NULL
)
@fix depends on !s@
position r.p1,r.p2;
expression x,E;
statement S1,S2;
@@
(
- if ((x@p2 != NULL) || ...)
S1
|
- if ((x@p2 == NULL) && ...) S1
|
- BUG_ON(x@p2 == NULL);
)
// </smpl>
Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Move the line disciplines towards a conventional ->ops arrangement. For
the moment the actual 'tty_ldisc' struct in the tty is kept as part of
the tty struct but this can then be changed if it turns out that when it
all settles down we want to refcount ldiscs separately to the tty.
Pull the ldisc code out of /proc and put it with our ldisc code.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Something Arjan suggested which allows us to clean up the code nicely
Signed-off-by: Alan Cox <alan@redhat.com>
Cc: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
- Operations are now a shared const function block as with most other Linux
objects
- Introduce wrappers for some optional functions to get consistent behaviour
- Wrap put_char which used to be patched by the tty layer
- Document which functions are needed/optional
- Make put_char report success/fail
- Cache the driver->ops pointer in the tty as tty->ops
- Remove various surplus lock calls we no longer need
- Remove proc_write method as noted by Alexey Dobriyan
- Introduce some missing sanity checks where certain driver/ldisc
combinations would oops as they didn't check needed methods were present
[akpm@linux-foundation.org: fix fs/compat_ioctl.c build]
[akpm@linux-foundation.org: fix isicom]
[akpm@linux-foundation.org: fix arch/ia64/hp/sim/simserial.c build]
[akpm@linux-foundation.org: fix kgdb]
Signed-off-by: Alan Cox <alan@redhat.com>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix a bunch of warnings in PPP and related drivers. Mostly because
sparse doesn't like it when the the function is only marked private in
the forward declaration.
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
PPP synchronous tty channel driver: convert the semaphore dead_sem to a
completion
Signed-off-by: Matthias Kaehlcke <matthias.kaehlcke@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
fix sparse warnings "Using plain integer as NULL pointer"
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
We conciously make a change here - we permit mode and speed setting to
be done in things like SLIP mode. There isn't actually a technical
reason to disallow this. It's usually a silly thing to do but we can
do it and soemone might wish to do so.
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Transform some calls to kmalloc/memset to a single kzalloc (or kcalloc).
Here is a short excerpt of the semantic patch performing
this transformation:
@@
type T2;
expression x;
identifier f,fld;
expression E;
expression E1,E2;
expression e1,e2,e3,y;
statement S;
@@
x =
- kmalloc
+ kzalloc
(E1,E2)
... when != \(x->fld=E;\|y=f(...,x,...);\|f(...,x,...);\|x=E;\|while(...) S\|for(e1;e2;e3) S\)
- memset((T2)x,0,E1);
@@
expression E1,E2,E3;
@@
- kzalloc(E1 * E2,E3)
+ kcalloc(E1,E2,E3)
[akpm@linux-foundation.org: get kcalloc args the right way around]
Signed-off-by: Yoann Padioleau <padator@wanadoo.fr>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Acked-by: Russell King <rmk@arm.linux.org.uk>
Cc: Bryan Wu <bryan.wu@analog.com>
Acked-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Dave Airlie <airlied@linux.ie>
Acked-by: Roland Dreier <rolandd@cisco.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Mauro Carvalho Chehab <mchehab@infradead.org>
Acked-by: Pierre Ossman <drzeus-list@drzeus.cx>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: "David S. Miller" <davem@davemloft.net>
Acked-by: Greg KH <greg@kroah.com>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
To clearly state the intent of copying from linear sk_buffs, _offset being a
overly long variant but interesting for the sake of saving some bytes.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
The API and code have been through various bits of initial review by
serial driver people but they definitely need to live somewhere for a
while so the unconverted drivers can get knocked into shape, existing
drivers that have been updated can be better tuned and bugs whacked out.
This replaces the tty flip buffers with kmalloc objects in rings. In the
normal situation for an IRQ driven serial port at typical speeds the
behaviour is pretty much the same, two buffers end up allocated and the
kernel cycles between them as before.
When there are delays or at high speed we now behave far better as the
buffer pool can grow a bit rather than lose characters. This also means
that we can operate at higher speeds reliably.
For drivers that receive characters in blocks (DMA based, USB and
especially virtualisation) the layer allows a lot of driver specific
code that works around the tty layer with private secondary queues to be
removed. The IBM folks need this sort of layer, the smart serial port
people do, the virtualisers do (because a virtualised tty typically
operates at infinite speed rather than emulating 9600 baud).
Finally many drivers had invalid and unsafe attempts to avoid buffer
overflows by directly invoking tty methods extracted out of the innards
of work queue structs. These are no longer needed and all go away. That
fixes various random hangs with serial ports on overflow.
The other change in here is to optimise the receive_room path that is
used by some callers. It turns out that only one ldisc uses receive room
except asa constant and it updates it far far less than the value is
read. We thus make it a variable not a function call.
I expect the code to contain bugs due to the size alone but I'll be
watching and squashing them and feeding out new patches as it goes.
Because the buffers now dynamically expand you should only run out of
buffering when the kernel runs out of memory for real. That means a lot of
the horrible hacks high performance drivers used to do just aren't needed any
more.
Description:
tty_insert_flip_char is an old API and continues to work as before, as does
tty_flip_buffer_push() [this is why many drivers dont need modification]. It
does now also return the number of chars inserted
There are also
tty_buffer_request_room(tty, len)
which asks for a buffer block of the length requested and returns the space
found. This improves efficiency with hardware that knows how much to
transfer.
and tty_insert_flip_string_flags(tty, str, flags, len)
to insert a string of characters and flags
For a smart interface the usual code is
len = tty_request_buffer_room(tty, amount_hardware_says);
tty_insert_flip_string(tty, buffer_from_card, len);
More description!
At the moment tty buffers are attached directly to the tty. This is causing a
lot of the problems related to tty layer locking, also problems at high speed
and also with bursty data (such as occurs in virtualised environments)
I'm working on ripping out the flip buffers and replacing them with a pool of
dynamically allocated buffers. This allows both for old style "byte I/O"
devices and also helps virtualisation and smart devices where large blocks of
data suddenely materialise and need storing.
So far so good. Lots of drivers reference tty->flip.*. Several of them also
call directly and unsafely into function pointers it provides. This will all
break. Most drivers can use tty_insert_flip_char which can be kept as an API
but others need more.
At the moment I've added the following interfaces, if people think more will
be needed now is a good time to say
int tty_buffer_request_room(tty, size)
Try and ensure at least size bytes are available, returns actual room (may be
zero). At the moment it just uses the flipbuf space but that will change.
Repeated calls without characters being added are not cumulative. (ie if you
call it with 1, 1, 1, and then 4 you'll have four characters of space. The
other functions will also try and grow buffers in future but this will be a
more efficient way when you know block sizes.
int tty_insert_flip_char(tty, ch, flag)
As before insert a character if there is room. Now returns 1 for success, 0
for failure.
int tty_insert_flip_string(tty, str, len)
Insert a block of non error characters. Returns the number inserted.
int tty_prepare_flip_string(tty, strptr, len)
Adjust the buffer to allow len characters to be added. Returns a buffer
pointer in strptr and the length available. This allows for hardware that
needs to use functions like insl or mencpy_fromio.
Signed-off-by: Alan Cox <alan@redhat.com>
Cc: Paul Fulghum <paulkf@microgate.com>
Signed-off-by: Hirokazu Takata <takata@linux-m32r.org>
Signed-off-by: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: John Hawkes <hawkes@sgi.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
This is part of the grand scheme to eliminate the qlen
member of skb_queue_head, and subsequently remove the
'list' member of sk_buff.
Most users of skb_queue_len() want to know if the queue is
empty or not, and that's trivially done with skb_queue_empty()
which doesn't use the skb_queue_head->qlen member and instead
uses the queue list emptyness as the test.
Signed-off-by: David S. Miller <davem@davemloft.net>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Gabriele Paoloni
Wei Yongjun
Julia Lawall
Alan Cox
Alan Cox
Stephen Hemminger
Matthias Kaehlcke
Joe Perches
Yoann Padioleau
Arnaldo Carvalho de Melo
Jeff Garzik
Arjan van de Ven
David S. Miller
Alexey Dobriyan
Linus Torvalds