wolf [Wed, 18 Feb 2009 18:38:12 +0000 (18:38 +0000)]
SJA1000: generic OF platform bus driver
This patch adds a generic driver for SJA1000 chips on the OpenFirmware
platform bus found on embedded PowerPC systems. You need a SJA1000 node
definition in your flattened device tree source (DTS) file similar to:
wolf [Tue, 17 Feb 2009 11:02:09 +0000 (11:02 +0000)]
Use EXPORT_SYMBOL_GPL instead of EXPORT_SYMBOL for drivers
As suggested by Sam, the safer approach is to be more restrictive.
If we later decide we want to support non-gpl we can be less restrictive.
Therefore let's go for the GPL version of the exports for now.
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@953 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Sun, 15 Feb 2009 17:40:32 +0000 (17:40 +0000)]
Spagetti code cleanup:
Moved conversions into separate functions can2pty() and pty2can().
Updated and enhanced some comments.
Variable name cleanups inside the separate functions.
Fix handling on wrong pty commands (was a wrong 'continue' statement).
wolf [Sun, 15 Feb 2009 10:38:25 +0000 (10:38 +0000)]
Prelimary quilt patch stack for including the device drivers interface
This is a prelimary quilt patch stack to include the CAN network
device driver interface and some drivers into Linux kernel. It is
based on the netdev GIT tree currently at version 2.6.29-rcX:
wolf [Sat, 14 Feb 2009 14:35:38 +0000 (14:35 +0000)]
Enable device dev_dbg with CONFIG_CAN_DEBUG_DEVICES
The following patch allows to enable CAN device debug messages via
CONFIG_CAN_DEBUG_DEVICES and adds some useful dev_dbg() to dev.c,
sja1000.c and mscan.c.
The old MSCAN driver used dev_dbg() the heavy way, e.g. printing
messages for each message sent and received. I think that's only
useful for development.
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@942 030b6a49-0b11-0410-94ab-b0dab22257f2
wolf [Sat, 14 Feb 2009 14:32:35 +0000 (14:32 +0000)]
MSCAN/MPC52xx: copyright note fixes and beautification
I restricted the copyright to GPL v2 and added a copyright line for myself.
Furthermore, I changed the file header to a more usual format. I also
changed the MODULE_AUTHOR in mpc52xx_can.c to myself for my re-written
version using the OF platform.
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@941 030b6a49-0b11-0410-94ab-b0dab22257f2
wolf [Sat, 14 Feb 2009 14:31:34 +0000 (14:31 +0000)]
dev.c: cleanup and name changes and more doc
This patch briefly documents exported functions of the CAN device
driver interface. I also changed the module description to "CAN
network device driver interface" because it's more appropriate than
"CAN netdevice library", I think. Comments?
Finally, I fix the issue marked with "#ifdef FIXME".
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@940 030b6a49-0b11-0410-94ab-b0dab22257f2
wolf [Sat, 14 Feb 2009 14:29:41 +0000 (14:29 +0000)]
drivers/net/can/Kconfig cleanup and beautification
This patch fixes some white space issues and uses consequently the tag
"---help---" for help text. Furthermore it changes the type of
CAN_CALC_BITTIMING to "bool".
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@938 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Fri, 6 Feb 2009 10:03:54 +0000 (10:03 +0000)]
Added CONFIG_CAN_CALC_BITTIMING to be defined by default when compiling
CAN drivers on the SVN trunk. This fixes an issue caused by commit r920:
"Make CAN bit-timing calculation configurable"
wolf [Sun, 1 Feb 2009 15:26:56 +0000 (15:26 +0000)]
SJA1000: remove unnecessary debugging code
For kernel inclusion, we need to cleanup and remove unnecessary
debugging code from the SJA1000 driver. Therefore I removed the
private DBG debugging interface and converted a few DBG calls to
dev_info() or dev_err(), where appropriate. Maybe they can be
removed as well, please check?
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@921 030b6a49-0b11-0410-94ab-b0dab22257f2
wolf [Sun, 1 Feb 2009 15:24:41 +0000 (15:24 +0000)]
Make CAN bit-timing calculation configurable
This patch adds the Kconfig option CAN_CALC_BITTIMING to make the
calculation of bit-timing parameters configurable. Disabling it will
save some space and will allow user space tools to determine proper
bit-timing parameters exclusively.
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@920 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Thu, 29 Jan 2009 17:04:44 +0000 (17:04 +0000)]
Fix 64bit issue in strtoul() usage reported by Kurt Van Dijck:
The idea is that the (unsigned long *) cast is wrong on 64bit,
therefore, a real unsigned long intermediate variable should be
applied. Before this patch, the can_dlc field was always cleared to 0
again.
Added in the patch is a line that I otherwise needed an ascii table to
understand what was going on. I took the freedom to make it more
programmer-readable wihthout performance penalty :-)
Signed-off-by: Kurt Van Dijck <kurt.van.dijck@eia.be> Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@919 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Wed, 21 Jan 2009 12:45:27 +0000 (12:45 +0000)]
Updated to latest CAN core changes:
From commit r890:
- removed avoidable copy of data in bcm_rx_handler()
- some removal of duplicate code
From commit r894:
Omit unneeded skb_clone() calls.
The AF_CAN core delivered always cloned sk_buffs to the AF_CAN
protocols, although this was _only_ needed by the can-raw protocol.
With this (additionally documented) change, the AF_CAN core calls the
callback functions of the registered AF_CAN protocols with the original
(uncloned) sk_buff pointer and let's the can-raw protocol do the
skb_clone() itself which omits all unneeded skb_clone() calls for other
AF_CAN protocols.
hartkopp [Wed, 21 Jan 2009 10:14:42 +0000 (10:14 +0000)]
Updated to latest CAN core changes:
From commit r890:
- removed avoidable copy of data in bcm_rx_handler()
- some removal of duplicate code
From commit r894:
Omit unneeded skb_clone() calls.
The AF_CAN core delivered always cloned sk_buffs to the AF_CAN
protocols, although this was _only_ needed by the can-raw protocol.
With this (additionally documented) change, the AF_CAN core calls the
callback functions of the registered AF_CAN protocols with the original
(uncloned) sk_buff pointer and let's the can-raw protocol do the
skb_clone() itself which omits all unneeded skb_clone() calls for other
AF_CAN protocols.
hartkopp [Wed, 21 Jan 2009 07:36:32 +0000 (07:36 +0000)]
Revert commit r910: "Added sanity check for received slcan ASCII data."
as this is done with ASCII comparision three lines above.
It obviously was too late yesterday ;-)
hartkopp [Tue, 20 Jan 2009 23:06:01 +0000 (23:06 +0000)]
Added a proof of concept tool 'slcanpty' which
creates a pty for applications using the slcan ASCII protocol and
converts the data to a CAN network interface (and vice versa).
This can be used for existing applications to run on SocketCAN.
wolf [Tue, 20 Jan 2009 18:51:50 +0000 (18:51 +0000)]
MSCAN-MPC52XX: re-write as proper OF driver (/arch/ppc is gone)
Starting with 2.6.28, the /arch/ppc tree is no longer available making
the legacy support obsolete. I took the occasion to re-write the driver
using the proper Open Firmware device driver interface.
Furthermore, MSCAN clock source selection has been enhanced. By default,
the external oscillator clock is now selected where appropriate (for
MPC5200B chips) because it's the better choice according to the MPC5200
user's manual as it has less jitter. The IP bus clock can be selected
via flat device tree property "clock-ipb":
wolf [Tue, 20 Jan 2009 18:50:21 +0000 (18:50 +0000)]
MSCAN: don't use interrrupt sharing
The MSCAN devices use dedicated interrupt sources and therefore
interrupt sharing is not needed. This patch corrects the relevant
flags of request_irq().
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@907 030b6a49-0b11-0410-94ab-b0dab22257f2
wolf [Tue, 20 Jan 2009 18:48:41 +0000 (18:48 +0000)]
dev: add register_candev() and unregister_candev()
Not all CAN devices are using the CAN device interface, like vcan or
slcan. Providing "struct can_priv" and SYSFS files for them seems
overkill. Because it's awkward to identify a device using the CAN
device interface in the netdev notifier callback, the functions
register_candev() and unregister_candev() have been introduced. All
devices using the CAN device interface should call them instead of
register_netdev() and unregister_netdev().
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
This line, and those below, will be ignored--
M kernel/2.6/include/linux/can/dev.h
M kernel/2.6/drivers/net/can/dev.c
M kernel/2.6/drivers/net/can/mcp251x.c
M kernel/2.6/drivers/net/can/mscan/mscan.c
M kernel/2.6/drivers/net/can/softing/softing_main.c
M kernel/2.6/drivers/net/can/sja1000/sja1000.c
hartkopp [Fri, 16 Jan 2009 11:51:44 +0000 (11:51 +0000)]
Fix crash on re-attaching slcan netdevices ...
... by removing the extra code added to slcan.c that was not in slip.c .
Indeed the slip driver (and so the slcan driver) holds the several
created slcan netdevices in a structure and only registeres the allocated
netdevices - but doesn't unregister them immediately at detach time.
Even when detached the devices still appear in /proc/net/dev (!)
Finally the destructor is well defined to execute free_netdev() at
unregister_netdev() which is handled properly.
The added free_netdev() produced a double free and crashed. Sorry.
hartkopp [Thu, 15 Jan 2009 11:14:12 +0000 (11:14 +0000)]
Reverted the removal of dev->get_stats = can_get_stats for kernels < 2.6.23
For these old kernels the access to the stats from outside the driver
only works via this function call.
hartkopp [Thu, 15 Jan 2009 08:30:18 +0000 (08:30 +0000)]
Update acces to netstats for 2.6.29
---
The successor of dev->get_stats(dev) in 2.6.29 is dev_get_stats(dev)
which provides a const(!) pointer for read only access to the stats.
As the usual current method to access the stats is to reference
dev->stats directly as stats is part of the struce netdevice
since 2.6.23 this patch implements this method of access now.
For kernels < 2.6.23 the netstats are part of the private data
and only this is handled by a separate wrapper now.
The idea was also to get a proper source code after strip-src.
hartkopp [Wed, 14 Jan 2009 08:02:23 +0000 (08:02 +0000)]
Due to the loopback functionality in can_send() we can not invoke it
from hardirq context which was done inside the bcm_tx_timeout_handler()
hrtimer callback:
hartkopp [Wed, 7 Jan 2009 08:38:59 +0000 (08:38 +0000)]
Added some sample code to use CAN interfaces with PF_PACKET sockets.
Just a proof-of-concept that has several drawbacks:
- you need to be root to access the CAN bus
- no local loopback (therefore virtual CAN interfaces do not really work)
- no CAN filters and other socket options from the can-raw socket
hartkopp [Tue, 6 Jan 2009 09:23:41 +0000 (09:23 +0000)]
Omit unneeded skb_clone() calls.
The AF_CAN core delivered always cloned sk_buffs to the AF_CAN
protocols, although this was _only_ needed by the can-raw protocol.
With this (additionally documented) change, the AF_CAN core calls the
callback functions of the registered AF_CAN protocols with the original
(uncloned) sk_buff pointer and let's the can-raw protocol do the
skb_clone() itself which omits all unneeded skb_clone() calls for other
AF_CAN protocols.
hartkopp [Sun, 28 Dec 2008 08:11:18 +0000 (08:11 +0000)]
Updated BCM to support hrtimers only run in hard_irq context in 2.6.29+
- added tasklets for messages to the userspace (needs soft_irq context)
- removed avoidable copy of data in bcm_rx_handler()
- fixed rearm of throttle timer in bcm_rx_thr_handler()
- some removal of duplicate code
hartkopp [Sun, 21 Dec 2008 16:55:48 +0000 (16:55 +0000)]
Created a copy of bcm.c for kernels below 2.6.22 (without hrtimers) to
split up the code as a precondition for changes caused by hrtimers in
2.6.29 that only support hrtimers in hard_irq context.
hartkopp [Thu, 18 Dec 2008 09:13:05 +0000 (09:13 +0000)]
can: merge error paths
This commit follows the change in net-next-2.6 git tree created by
Ilpo Järvinen.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
http://git.kernel.org/?p=linux/kernel/git/davem/net-next-2.6.git;a=commitdiff;h=ebad5c0984e06f4036d2eb2741b9afce736b8a06
hartkopp [Mon, 8 Dec 2008 18:51:29 +0000 (18:51 +0000)]
Fix NULL pointer dereference as reported by Chris Dore:
https://lists.berlios.de/pipermail/socketcan-core/2008-December/002120.html
Added missing initialisation of resource struct as reported by David Mueller:
https://lists.berlios.de/pipermail/socketcan-core/2008-December/002127.html
Signed-off-by: David Mueller <d.mueller@elsoft.ch> Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@879 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Thu, 4 Dec 2008 18:21:43 +0000 (18:21 +0000)]
Since rev872 the single CAN identifier filter lists handle only non-RTR
CAN frames. So we need to omit the check of these filter lists when
receiving RTR CAN frames.
hartkopp [Mon, 1 Dec 2008 07:24:18 +0000 (07:24 +0000)]
Fix sloppy CAN_(EFF|RTR)_FLAG handling in can_filter.can_mask .
Due to a wrong safety check in af_can.c it was not possible to filter
for SFF frames with a specific CAN identifier without getting the
same selected CAN identifier from a received EFF frame also.
This fix has a minimum impact on the CAN filter API as the 'sloppy'
handling is still a correct (and possibly wanted?) use-case.
Please update the can-utils (especially candump) whose filter definition
on the commandline made assumptions to correct the user input that are
probably unwanted now.
Thanks to Kurt van Dijck for pointing at this issue!
Signed-Off-by: Oliver Hartkopp <oliver@hartkopp.net>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@872 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Fri, 28 Nov 2008 21:50:24 +0000 (21:50 +0000)]
Fixed parse_canframe() with reordering the checks for the can_id
delimiter. Before this fix the function may have found delimiters in the
string behind the terminating zero.
hartkopp [Thu, 30 Oct 2008 21:18:34 +0000 (21:18 +0000)]
[PATCH] SJA1000: ixxat_pci: fix the detection of the second SJA1000 chip
According to the SJA1000 data sheet, section 6.3.2, only the bits 0,5
and 7 of the control/mode register are defined after a reset. This
patch uses a proper mask before checking the value.
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com> Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@863 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Wed, 29 Oct 2008 19:05:06 +0000 (19:05 +0000)]
This patch allows pcmcia subsystem >= 2.6.28
The function pcmcia_loop_config is used. To support kernels < 2.6.28
the pcmcia_loop_config function is emulated (it does basically the same
as what I used to do, but seperates the code much cleaner) for these kernels.
This is definitely a change, but does not break on recent builds.
Additionally, in order to test the pcmcia access, the code needed to be
tested for 8bit softing pcmcia cards too. Historically, these cards did not
support shared irq's. But my current system, AMD64, does not provide exclusive
interrupts anymore. Therefore, the old 8bit cards are moved to have shared
irq's. The code did succeed in my elementary tests.
Signed-off-by: Kurt Van Dijck <kurt.van.dijck@eia.be> Signed-off-by: Oliver Hartkopp <oliver@hartkopp.net>
git-svn-id: svn://svn.berlios.de//socketcan/trunk@862 030b6a49-0b11-0410-94ab-b0dab22257f2
hartkopp [Tue, 28 Oct 2008 19:20:02 +0000 (19:20 +0000)]
Added a file list and base patches to create a Kernel 2.6.25+ with the latest CAN divers.
The base patches just remove the vcan driver files which are taken from SVN also.
projects / socketcan-devel.git / log