Fix typo in atmsar.c
Moved the version number into the speedtouch source file.
Added GPL licence notice to source files. Cleaned up whitespace and changelogs.
Correct wierd module init and cleanup style.
initial checkin
Fixed page rot. Added myself as new maintainer.
Add GPL module licence.
Include slab.h rather than malloc.h
Change urb_t to struct urb.
Completed the integration of sarlib into the kernel module.
Remove assumption about processor type.
Cause make install to run depmod.
Backport whitespace cleanups from 2.5 kernel version.
Replace deprecated FILL_BULK_URB by usb_fill_bulk_urb (backport of 2.5
Add mutex to serialise calls to the usbdevfs ioctl handler (backport of 2.5
Remove unneeded inclusion of version.h.
Merged the changelog into the source file (as in the 2.5 kernels).
Added more information about getting hold of an ATM enabled PPP daemon.
Micro whitespace backport from 2.5.
Fix unlinking race. Backport of 2.5 patch by oliver@neukum.name
Bundled sarlib with the kernel module, as in the 2.5 kernels.
Remove call to usb_clear_halt (which may sleep) from completion handler.
Simplify spinlocks: no need to turn off interrupts in an irq handler (backport
Remove broken code from the disconnect handler (backport of 2.5 patch by
Fix possible deadlock in atm_close path (backport of 2.5 patch by
Updated and fleshed out (a little) the installation instructions.
daemonize does exit_files (backport from 2.5; maybe due to Richard Purdie?)
Bumped version to 1.5A; credited Richard Purdie.
Rolled udsl_atm_set_mac into udsl_usb_probe.
Replace the kernel thread with a tasklet.
Get rid of pointless casts from void *.
Rework error handling in udsl_usb_probe.
Miscellaneous backports from 2.5 (maybe due to Richard Purdie?)
Revert FILL_BULK_URB -> usb_fill_bulk_urb change for 2.4.15 compatibility.
struct sk_buff_head has a spinlock built in, so no need for another one.
Kill off tasklet on shutdown. Race pointed out by Oliver Neukum <oliver@neukum.name>
Make more functions static.
Fix race due to bad design of atm_dev_register.
Note that kernel 2.4.20 or better is needed for the version in cvs.
Rolled udsl_atm_startdevice into udsl_usb_probe.
Eliminate sleep_on (backport of 2.5 patch by oliver@neukum.name)
Add note about race after usb_unlink_urb.
Export usb device info using MODULE_DEVICE_TABLE. This may cause problems
Forgot to remove the old receive queue of sk_bufs!
Removed minor table. This means that there is no longer any limit
Do not use a dynamically allocated array for rcvbufs.
Remove unused printk arguments.
Added Freeserve (UK) VPI/VCI info.
Trivial cleanups (whitespace etc).
Whitespace clean up. Only use "SpeedTouch" (was sometimes "Speed Touch").
Forgot part of previous change: dynamic allocation of send urbs. In fact this
Pass the receive urb to the tasklet rather than just the sk_buff.
Clean kernel thread exit (backport of 2.5 patch by oliver@neukum.name)
Simplify the receive urb lifecycle: allocation in probe, deallocation in disconnect.
Revert FILL_BULK_URB -> usb_fill_bulk_urb change for 2.4.15 compatibility.
Dynamically allocate send urbs (backport of 2.5 patch by Richard Purdie).
Simplify debug flag handling.
Added a new encoding routine atmsar_encode which combines AAL5 frame
Change udsl_process_receive to not allocate a sk_buff on the stack.
Revert comment change caused by overzealous search and replace.
Deal with usb_unlink_urb failures in disconnect path.
Added VCI/VPI for New Zealand. Sorted the list into alphabetical order.
Allow ATM devices to be opened before the firmware is uploaded. This is the
Added link to the SuSe speedtouch webpage. Was more explicit about where
Correct atmsar_encode debug message.
Reuse receive urb buffers (now never reallocated).
Fix the simplified list handling in udsl_fire_receivers.
Removed passive option from ppp/peers files.
Deal better with receive urb failures.
The previous patch unlocked the wrong spinlock!
Added persist option to ppp peers file.
Change data_started to firmware_loaded. Get rid of udsl_usb_data_exit.
Cosmetic changes.
Refuse to open a VCC if the firmware is not loaded. This is not backwards compatible.
Return -EAGAIN rather than -ENODEV when udsl_atm_open is called before the firmware is loaded.
Allocate send urbs in udsl_usb_probe. Since this diminishes
Added VPI/VCI for Telstra, Australia.
Simplify list handling in udsl_fire_receivers.
More tweaks to the ppp/peers files.
Wait for receive completion handlers to finish after usb_unlink_urb.
Didn't recompile if atmsar.h was changed.
Use alternate setting 1 on interface 1 (was alternate 2). This should mean
Simplify failure handling in udsl_usb_probe.
Deal more strictly with a failure to get the serial number.
Don't access memory after freeing it! Plus some formating changes.
Clean up hidden files created by the kernel make process.
Keep a reference to the USB device until the ATM layer goes down (because
It seems Telstra uses PPPoE.
Put in the infrastructure for the new send code.
Don't extract USB information in udsl_atm_proc_read.
Measure the receive buffer size in ATM cells (53 bytes).
Makefile didn't recompile when a source file changed!
Don't crash on shutdown. The driver should now be usable.
Forgot to update the stats transmission counter! Also, let's not pretend we
Replace PDEBUG with USB dbg macro.
Fix debugging printk
Publish correct device name and ADSL line status in /proc/net/atm
Suggest cvs version rather than last (very out of date) release.
Fix memory leak found by the Stanford Checker (patch by Randy Dunlap).
The new send code. Enjoy!