The sealevel driver called sppp_attach before checking that board existed
and never called detach in the error path.

My change is to call sppp_attach from the netdev->init hook which happens
later in the process, and call detach from the uninit hook.

Also, changed the structure element 'netdev' to 'pppdev' to avoid confusion.

Here is the fix against 2.6.0-test5 latest

For e1000 and baycom_epp.

Contributed by Jeff Garzik.

Hi,
Please apply to 2.6.0-test5-current.

Thanks,
--
~Randy


From: Randy Hron <rwhron@earthlink.net>


Remove unneeded include of version.h.
Test compiled against 2.6.0-test5-bk9.


 linux-260-t5bk12-kj-rddunlap/drivers/net/e100/e100.h |    1 -
 1 files changed, 1 deletion(-)

Hi,
Please apply to 2.6.0-test5-current.

Thanks,
--
~Randy


From: Domen Puncer <domen@coderock.org>

IMO, that verify_area wasn't needed.



 linux-260-t5bk12-kj-rddunlap/drivers/net/wireless/orinoco.c |    4 ----
 1 files changed, 4 deletions(-)

* Some archs don't like calling pci_alloc_consistent before calling
pci_enable_device.  (Imagine that).  This corrects that.

* Driver was advertising HW_CSUM, but hardware is only capable of IP_CSUM.

* PRO/10+ (10 Mbps-only card) was not configured properly so it didn't
pass traffic.

* removed unused var, ASSERT macro
* missed a free_netdev() in probe cleanup undo.

* Better propagation of error codes during probe/open paths.
  {Janice Girouard (janiceg@us.ibm.com)]

* Bug fix: SERDES devices might be connected to back-plan switch that
  doesn't support auto-neg, so add the capability to force 1000/full.

* handle ethtool force flow control
* correctly set flow control hi/low watermarks based on size of Rx FIFO
  area.  The size can change if doing Jumbo Frames or, in the case of
  82547, is smaller to start with.
* was not properly forcing flow control settings to fc_none if using
  strict IEEE flow control override.

Replace proc_read with seq_file interface which is much cleaner.

Also, instead of searching the list of devices looking for the
name which is slow and would break if interface name changes;
store a pointer to device in the proc entry, and retrieve
it from the information saved by single_open.

Formatting and other behaviours are retained.

Existing code does not
	- check if proc_fs functions return error.
	- use proc_mkdir to make the directory.

Note pSkRootDir defined twice.

There is no exported variable called smp_num_cpus! so this driver
won't build on SMP.  Since the local variable is never used anyway
just get rid of it.

This applies against 2.6.0-test5-bk13 which has the last vendor
driver update.

*duh*

Use "select CRC32" in Kconfig instead of makefile includes.

into home.osdl.org:/home/torvalds/v2.5/linux

into home.osdl.org:/home/torvalds/v2.5/linux

into home.osdl.org:/home/torvalds/v2.5/linux

This is from Ben Herrenschmidt's tree.

into samba.org:/home/paulus/kernel/for-linus-ppc

into samba.org:/home/paulus/kernel/for-linus-ppc

into samba.org:/home/paulus/kernel/for-linus-ppc

This moves the uc_sigmask field back to where it was before.  We now don't
assume the registers are at a fixed offset in the ucontext, but instead
access them through the uc_regs field as in the past.  This corresponds with
a recent change in glibc CVS.

From: Douglas Gilbert <dougg@torque.net>

Attached is an update of the scsi_mid_low_api text
document. The first attachment is a gzipped patch against
lk 2.6.0-test4 and the second one is the file gzipped.

Changes:
   - add "Calling context:" entry to interface functions
   - add "Might block:" entry to mid level functions
   - drop scsi_set_device_offline() + command()
   - change references to reflect transfer of headers to
     include/scsi/scsi_*.h
   - try to define what a "scsi host" is
   - typos + grammar

Thanks to Randy Dunlap, Alan Stern amd Christoph Hellwig
for their input. Comments welcome.

into raven.il.steeleye.com:/home/jejb/BK/scsi-for-linus-2.6

GCC 3.4 miscompiles the kernel because it silently optimizes away
data placed in the .init.setup section by the __setup() macro.

__attribute__((unused)) does only avoid the warning, but doesn't
mark the data as being used.

Since GCC 3.3, __attribute__((used)) should be applied to such
variables.  The __attribute_used__ macro from linux/compiler.h already
takes care of compiler differences for us.

In this patch, I've gone a step further and proactively fixed that in
all places.

Thanks Matt and Jens for the debug help on the multipath problem. 

This solves the problem and makes multipath work properly.

There are two types of "flags" that are used in a block io request,
bi_flags, and bi_rw.  bi_flags is used for flags to the block level
code, and bi_rw is used for flags to the low level device drivers.

The code in the multipath driver used the wrong flag in the wrong field.
In this case, the flag FASTFAIL (value 3) was being set to the bi_flags
field.  FASTFAIL is a hint to the low level driver that it should try to
fail out quickly.  Unfortunately, the value 3 is also BIO_SEG_VALID,
which is a flag to the block subsystem that the segments shouldn't be
recalculated.  The result was that the wrong field was set, telling the
block layer not to recalculate the segments resulting in phys and hw
segments of 0.  Not good.

This is a new patch that includes Martin's, and makes create_dir() return
an int. That matches sysfs_create_file(), and saves a couple of
conversions back and forth between pointer and error.

It's not like it would compile without PCI support anyway, and these
ifdefs are _really_ strange.

Same issues as the dma mapping code, but as SN2 interrupt handling is
totally static and all irqs are allocated on bootup an never freed it's
not a speed issue.

I really wonder what someone has to smoke to call a that expensive
dispatcher just to cast the data structure back to the lowlevel one
directly afterwards...

fed through Lindent and dead codepathes eliminated.

Currently the the SN2 dma mapping code uses a file systen lookup (cat /r/napali/tmp/p10 |diffstat -p1) to
find the operation vector for the underlying dma mapping implementation.

But not only there is exactly one of those implementation, also data
structures from this lowlevel driver are used in the actual Linux dma
mapping routines and quite a few assumption are made about it.  So let's
just call directly in the lowlevel code and give up this utter bullshit.

These strange mixups of Linux and IRIX PCI interfaces were used in SGI's
2.4 tree for their port of the IRIX scsi code.  They have no business to
stgay in 2.6.