openwrt/staging/blogic.git
19 years ago[PATCH] aio: clean up debug code
Ken Chen [Sun, 1 May 2005 15:59:15 +0000 (08:59 -0700)]
[PATCH] aio: clean up debug code

Clean up code that was previously used for debug purpose.  Remove aio_run,
aio_wakeups, iocb->ki_queued and iocb->ki_kicked.  Also clean up unused
variable count in __aio_run_iocbs() and debug code in read_events().

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Suparna Bhattacharya <suparna@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] aio: ring wrapping simplification
Ken Chen [Sun, 1 May 2005 15:59:15 +0000 (08:59 -0700)]
[PATCH] aio: ring wrapping simplification

Since the tail pointer in aio_ring structure never wrap ring size more than
once, so a simple compare is sufficient to wrap the index around.  This avoid
a more expensive mod operation.

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Suparna Bhattacharya <suparna@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] aio: remove superfluous kiocb member initialization
Ken Chen [Sun, 1 May 2005 15:59:15 +0000 (08:59 -0700)]
[PATCH] aio: remove superfluous kiocb member initialization

This patch removes superfluous kiocb member initialization in the AIO
allocation and deallocation path.  For example, in really_put_req(),
right before kiocb is returned to slab, 5 variables are reset to NULL.
The same variables will be initialized at the kiocb allocation time,
so why bother reset them knowing that they will be set to valid data
at alloc time?  Another example: ki_retry is initialized in __aio_get_req,
but is initialized again in io_submit_one.

Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Cc: Suparna Bhattacharya <suparna@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Makefile: fix for compatibility with *emacs ctags
Emanuele Giaquinta [Sun, 1 May 2005 15:59:15 +0000 (08:59 -0700)]
[PATCH] Makefile: fix for compatibility with *emacs ctags

I've noticed that, starting from linux-2.6.12-rc1, in the top Makefile the
"cmd_tags" variable has been changed in a way incompatible with *emacs
ctags.  Since the "--extra" option exists only in "exuberant ctags", it
should be included in the CTAGSF shell variable.

Cc: Sam Ravnborg <sam@ravnborg.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] serial_cs: Reduce stack usage in serial_event()
Yum Rayan [Sun, 1 May 2005 15:59:14 +0000 (08:59 -0700)]
[PATCH] serial_cs: Reduce stack usage in serial_event()

This patch reduces the stack usage of the function serial_event() in
serial_cs from 2212 to 228.  I used a patched version of gcc 3.4.3 on i386
with -fno-unit-at-a-time disabled.

This patch is only compile tested.

Acked-by: Randy Dunlap <rddunlap@osdl.org>
Signed-off-by: Yum Rayan <yum.rayan@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] fix include order in mthca_memfree.c
Roland Dreier [Sun, 1 May 2005 15:59:14 +0000 (08:59 -0700)]
[PATCH] fix include order in mthca_memfree.c

Fix order of #include lines in mthca_memfree.c

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] convert that currently tests _NSIG directly to use valid_signal()
Jesper Juhl [Sun, 1 May 2005 15:59:14 +0000 (08:59 -0700)]
[PATCH] convert that currently tests _NSIG directly to use valid_signal()

Convert most of the current code that uses _NSIG directly to instead use
valid_signal().  This avoids gcc -W warnings and off-by-one errors.

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] new valid_signal() function
Jesper Juhl [Sun, 1 May 2005 15:59:13 +0000 (08:59 -0700)]
[PATCH] new valid_signal() function

This patch adds a new function valid_signal() that tests if its argument is
a valid signal number.

The reasons for adding this new function are:

- some code currently testing _NSIG directly has off-by-one errors.
  Using this function instead avoids such errors.

- some code currently tests unsigned signal numbers for <0 which is
  pointless and generates warnings when building with gcc -W.  Using this
  function instead avoids such warnings.

I considered various places to add this function but eventually settled on
include/linux/signal.h as the most logical place for it.  If there's some
reason this is a bad choice then please let me know (hints as to a better
location are then welcome of course).

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Fix Tpm driver -- Maintainers entry
Kylene Hall [Sun, 1 May 2005 15:59:13 +0000 (08:59 -0700)]
[PATCH] Fix Tpm driver -- Maintainers entry

This patch adds the maintainers entry.

Signed-off-by: Kylene Hall <kjhall@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] consolidate sys_shmat
Stephen Rothwell [Sun, 1 May 2005 15:59:12 +0000 (08:59 -0700)]
[PATCH] consolidate sys_shmat

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] sn_console: make sal_console_uart static again
Andreas Schwab [Sun, 1 May 2005 15:59:12 +0000 (08:59 -0700)]
[PATCH] sn_console: make sal_console_uart static again

Signed-off-by: Andreas Schwab <schwab@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ipmi: fix a deadlock
Corey Minyard [Sun, 1 May 2005 15:59:12 +0000 (08:59 -0700)]
[PATCH] ipmi: fix a deadlock

Correct an issue with the IPMI message layer taking a lock and calling
lower layer driver.  If an error occrues at the lower layer the lock can be
taken again causing a deadlock.  The lock is released before calling the
lower layer.

Signed-off-by: David Griego <dgriego@mvista.com>
Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ipmi: enable interrupts on the BT driver
Corey Minyard [Sun, 1 May 2005 15:59:11 +0000 (08:59 -0700)]
[PATCH] ipmi: enable interrupts on the BT driver

Enable interrupts for a BT interface.  There is a specific register that
needs to be set up to enable interrupts that also must be modified to clear
the irq.

Also, don't reset the BMC on a BT interface.  That's probably not a good
idea as the BMC may be performing other important functions and a reset
should only be a last resort.  Also, that register is also used to
enable/disable interrupts to the BT; modifying it may screw up the
interrupts.

Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ipmi: fix watchdog so the device can be reopened on an unexpected close
Corey Minyard [Sun, 1 May 2005 15:59:11 +0000 (08:59 -0700)]
[PATCH] ipmi: fix watchdog so the device can be reopened on an unexpected close

If there is an unexpected close, still allow the watchdog interface to be
re-opened on the IPMI watchdog.

Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IPMI: fix for handling bad ACPI data
Corey Minyard [Sun, 1 May 2005 15:59:10 +0000 (08:59 -0700)]
[PATCH] IPMI: fix for handling bad ACPI data

If the ACPI register bit width is zero (an invalid value) assume it is the
default spacing.  This avoids some coredumps on invalid data and makes some
systems work that have broken ACPI data.

Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IPMI: fix for handling bad IPMI DMI data
Corey Minyard [Sun, 1 May 2005 15:59:10 +0000 (08:59 -0700)]
[PATCH] IPMI: fix for handling bad IPMI DMI data

Ignore the bottom bit of the base address from the DMI data.  It is
supposed to be set to 1 if it is I/O space.  Few systems do this, but this
enables the ones that do set it to work properly.

Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] dontdiff file sorted in alphabet order
aquynh@gmail.com [Sun, 1 May 2005 15:59:10 +0000 (08:59 -0700)]
[PATCH] dontdiff file sorted in alphabet order

Documentation/dontdiff is a little messy.  Here is a patch to sort the
content of that file in alphabetical

Signed-off-by: Nguyen Anh Quynh <aquynh@gmail.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] reiserfs: journal_init fix
Edward Shishkin [Sun, 1 May 2005 15:59:09 +0000 (08:59 -0700)]
[PATCH] reiserfs: journal_init fix

This fixes segmentation fault when specifying bad journal device via
a mount option.

Don't pass a zero pointer to bdevname() if filp_open() returns error.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Leadtek Winfast remote controls
Nicolas Boichat [Sun, 1 May 2005 15:59:09 +0000 (08:59 -0700)]
[PATCH] Leadtek Winfast remote controls

Add missing button codes for the Leadtek Winfast remote controls.

Signed-off-by: Nicolas Boichat <nicolas@boichat.ch>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] __attribute__ placement fixes
Vinay K Nallamothu [Sun, 1 May 2005 15:59:09 +0000 (08:59 -0700)]
[PATCH] __attribute__ placement fixes

The variable attributes "packed" and "align" when used with struct, should
have the following order:

struct ... {...} __attribute__((packed)) var;

This patch fixes few instances where the variable and attributes are placed
the other way around and had no effect.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] misc verify_area cleanups
Jesper Juhl [Sun, 1 May 2005 15:59:08 +0000 (08:59 -0700)]
[PATCH] misc verify_area cleanups

There were still a few comments left refering to verify_area, and two
functions, verify_area_skas & verify_area_tt that just wrap corresponding
access_ok_skas & access_ok_tt functions, just like verify_area does for
access_ok - deprecate those.

There was also a few places that still used verify_area in commented-out
code, fix those up to use access_ok.

After applying this one there should not be anything left but finally
removing verify_area completely, which will happen after a kernel release
or two.

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] consolidate SIGEV_PAD_SIZE
Stephen Rothwell [Sun, 1 May 2005 15:59:08 +0000 (08:59 -0700)]
[PATCH] consolidate SIGEV_PAD_SIZE

Discussing with Matthew Wilcox some of his outstanding patches lead me to
this patch (among others).

The preamble in struct sigevent can be expressed independently of the
architecture.

Also use __ARCH_SI_PREAMBLE_SIZE on ia64.

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] hangcheck-timer: Update to 0.9.0.
Joel Becker [Sun, 1 May 2005 15:59:08 +0000 (08:59 -0700)]
[PATCH] hangcheck-timer: Update to 0.9.0.

I recently realized that the in-kernel copy of hangcheck-timer was quite
stale.  Here's the latest.  It adds support for s390, ppc64, and ia64 too.

Signed-off-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] vgacon: set vc_hi_font_mask correctly
Bill Nottingham [Sun, 1 May 2005 15:59:07 +0000 (08:59 -0700)]
[PATCH] vgacon: set vc_hi_font_mask correctly

When allocating a new VC with vgacon_init(), the font is shared across all
the VGA consoles.  However, the font mask was always set to the default
value of zero in visual_init(), even if we were using 512 character fonts
at the time.

Moreover, code in vgacon.c:vga_do_font_op() didn't reset the mask if the
console driver thinks it's already in 512 character mode.  This means that
to *fix* it, you'd actually have to take the console out of 512 character
mode and then set it back.

The attached sets vc_hi_font_mask in vgacon_init() for any new consoles
opened if the vgacon driver is already in 512 character mode, solving this.

This bug goes back to 2.4.18 at least, probably earlier.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Fix rewriting on a full reiserfs filesystem
Jan Kara [Sun, 1 May 2005 15:59:07 +0000 (08:59 -0700)]
[PATCH] Fix rewriting on a full reiserfs filesystem

Allow rewriting of a file and extending a file upto the end of the
allocated block on a full filesystem.

From: Chris Mason <mason@suse.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] nbd: Don't create all MAX_NBD devices by default all the time
Lars Marowsky-Bree [Sun, 1 May 2005 15:59:07 +0000 (08:59 -0700)]
[PATCH] nbd: Don't create all MAX_NBD devices by default all the time

This patches adds the "nbds_max" parameter to the nbd kernel module, which
limits the number of nbds allocated.  Previously, always all 128 entries
were allocated unconditionally, which used to waste resources and
needlessly flood the hotplug system with events.  (Defaults to 16 now.)

Signed-off-by: Lars Marowsky-Bree <lmb@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] add EOWNERDEAD and ENOTRECOVERABLE version 2
Joe Korty [Sun, 1 May 2005 15:59:06 +0000 (08:59 -0700)]
[PATCH] add EOWNERDEAD and ENOTRECOVERABLE version 2

Add EOWNERDEAD and ENOTRECOVERABLE to all architectures.  This is to
support the upcoming patches for robust mutexes.

We normally don't reserve parts of the name/number space for external
patches, but robust mutexes are sufficiently popular and important to
justify it in this case.

Signed-off-by: Joe Korty <joe.korty@ccur.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] noop-iosched: kill O(N) merge scan
Jens Axboe [Sun, 1 May 2005 15:59:06 +0000 (08:59 -0700)]
[PATCH] noop-iosched: kill O(N) merge scan

Profiling hit rates on merging shows that the last merge hint works
extremely well for most work loads.  So lets kill the linear merge scan in
noop-iosched, so it provides O(1) run time for any operation.

Testing credits go to Ken Chen from Intel.

Signed-off-by: Jens Axboe <axboe@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] kallsyms C_SYMBOL_PREFIX support
Yoshinori Sato [Sun, 1 May 2005 15:59:06 +0000 (08:59 -0700)]
[PATCH] kallsyms C_SYMBOL_PREFIX support

kallsyms does not consider SYMBOL_PREFIX of C.  Consequently it does not
work on architectures using that prefix character (h8300, v850).

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] LifeView FlyTV Platinum FM: GPIO usage
Peter Missel [Sun, 1 May 2005 15:59:05 +0000 (08:59 -0700)]
[PATCH] LifeView FlyTV Platinum FM: GPIO usage

This is take two of a patch that should have appeared two days ago, before
yesterday's "remote control" patch for the same card.

This patch sets unconnected GPIO to Output to keep them from floating (just
good driver writing practice, being nice to the chip), and uses GPIO16 to
switch TV vs.  FM - this pin switches inputs onto the tuner, as well as the
audio output from the tuner into the 7135 SIF input.  Consequently, FM
radio support is being un-commented because it's now working (sort of, see
below).

These two patches get the card almost fully operational; there appears to
be a bug in tda8290.c remaining that puts an offset onto the tuned
frequency in FM radio mode.  We're investigating.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] LifeView FlyTV Platinum FM: Remote Control support
Peter Missel [Sun, 1 May 2005 15:59:05 +0000 (08:59 -0700)]
[PATCH] LifeView FlyTV Platinum FM: Remote Control support

Subject says it ...  this card's IR microcontroller design and attachment
are compatible to the company's previous designs, so the patch was as
simple as it gets.
DESC
LifeView FlyTV Platinum FM: GPIO usage
EDESC
From: Peter Missel <peter.missel@onlinehome.de>

This is take two of a patch that should have appeared two days ago, before
yesterday's "remote control" patch for the same card.

This patch sets unconnected GPIO to Output to keep them from floating (just
good driver writing practice, being nice to the chip), and uses GPIO16 to
switch TV vs.  FM - this pin switches inputs onto the tuner, as well as the
audio output from the tuner into the 7135 SIF input.  Consequently, FM
radio support is being un-commented because it's now working (sort of, see
below).

These two patches get the card almost fully operational; there appears to
be a bug in tda8290.c remaining that puts an offset onto the tuned
frequency in FM radio mode.  We're investigating.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] reiserfs: make resize option auto-get new device size
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:59:05 +0000 (08:59 -0700)]
[PATCH] reiserfs: make resize option auto-get new device size

It's trivial for the resize option to auto-get the underlying device size,
while it's harder for the user.  I've copied the code from jfs.

Since of the different reiserfs option parser (which does not use the
superior match_token used by almost every other filesystem), I've had to
use the "resize=auto" and not "resize" option to specify this behaviour.
Changing the option parser to the kernel one wouldn't be bad but I've no
time to do this cleanup in this moment.

Btw, the mount(8) man page should be updated to include this option.  Cc
the relevant people, please (I hope I cc'ed the right people).

Cc: <reiserfs-dev@namesys.com>
Cc: <reiserfs-list@namesys.com>
Cc: <mtk-manpages@gmx.net>
Cc: Alex Zarochentsev <zam@namesys.com>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Update RCU documentation
Paul E. McKenney [Sun, 1 May 2005 15:59:05 +0000 (08:59 -0700)]
[PATCH] Update RCU documentation

Update the RCU documentation to allow for the new synchronize_rcu() and
synchronize_sched() primitives.  Fix a few other nits as well.

Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Change synchronize_kernel to _rcu and _sched
Paul E. McKenney [Sun, 1 May 2005 15:59:04 +0000 (08:59 -0700)]
[PATCH] Change synchronize_kernel to _rcu and _sched

This patch changes calls to synchronize_kernel(), deprecated in the earlier
"Deprecate synchronize_kernel, GPL replacement" patch to instead call the new
synchronize_rcu() and synchronize_sched() APIs.

Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Deprecate synchronize_kernel, GPL replacement
Paul E. McKenney [Sun, 1 May 2005 15:59:04 +0000 (08:59 -0700)]
[PATCH] Deprecate synchronize_kernel, GPL replacement

The synchronize_kernel() primitive is used for quite a few different purposes:
waiting for RCU readers, waiting for NMIs, waiting for interrupts, and so on.
This makes RCU code harder to read, since synchronize_kernel() might or might
not have matching rcu_read_lock()s.  This patch creates a new
synchronize_rcu() that is to be used for RCU readers and a new
synchronize_sched() that is used for the rest.  These two new primitives
currently have the same implementation, but this is might well change with
additional real-time support.  Both new primitives are GPL-only, the old
primitive is deprecated.

Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Add deprecated_for_modules
Paul E. McKenney [Sun, 1 May 2005 15:59:03 +0000 (08:59 -0700)]
[PATCH] Add deprecated_for_modules

Add a deprecated_for_modules macro that allows symbols to be deprecated only
when used by modules, as suggested by Andrew Morton some months back.

Signed-off-by: Paul E. McKenney <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] kernel/rcupdate.c: make the exports EXPORT_SYMBOL_GPL
Paul E. McKenney [Sun, 1 May 2005 15:59:03 +0000 (08:59 -0700)]
[PATCH] kernel/rcupdate.c: make the exports EXPORT_SYMBOL_GPL

The gpl exports need to be put back.  Moving them to GPL -- but in a
measured manner, as I proposed on this list some months ago -- is fine.
Changing these particular exports precipitously is most definitely -not-
fine.  Here is my earlier proposal:

http://marc.theaimsgroup.com/?l=linux-kernel&m=110520930301813&w=2

See below for a patch that puts the exports back, along with an updated
version of my earlier patch that starts the process of moving them to GPL.
I will also be following this message with RFC patches that introduce two
(EXPORT_SYMBOL_GPL) interfaces to replace synchronize_kernel(), which then
becomes deprecated.

Signed-off-by: <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] procfs: Fix hardlink counts for /proc/<PID>/task
Daniel Drake [Sun, 1 May 2005 15:59:03 +0000 (08:59 -0700)]
[PATCH] procfs: Fix hardlink counts for /proc/<PID>/task

The current logic assumes that a /proc/<PID>/task directory should have a
hardlink count of 3, probably counting ".", "..", and a directory for a
single child task.

It's fairly obvious that this doesn't work out correctly when a PID has
more than one child task, which is quite often the case.

Signed-off-by: Daniel Drake <dsd@gentoo.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] procfs: Fix hardlink counts
Daniel Drake [Sun, 1 May 2005 15:59:03 +0000 (08:59 -0700)]
[PATCH] procfs: Fix hardlink counts

The pid directories in /proc/ currently return the wrong hardlink count - 3,
when there are actually 4 : ".", "..", "fd", and "task".

This is easy to notice using find(1):
cd /proc/<pid>
find

In the output, you'll see a message similar to:

find: WARNING: Hard link count is wrong for .: this may be a bug in your
filesystem driver.  Automatically turning on find's -noleaf option.
Earlier results may have failed to include directories that should have
been searched.

http://bugs.gentoo.org/show_bug.cgi?id=86031

I also noticed that CONFIG_SECURITY can add a 5th: attr, and performed a
similar fix on the task directories too.

Signed-off-by: Daniel Drake <dsd@gentoo.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] move SA_xxx defines to linux/signal.h
Stas Sergeev [Sun, 1 May 2005 15:59:02 +0000 (08:59 -0700)]
[PATCH] move SA_xxx defines to linux/signal.h

The attached patch moves the IRQ-related SA_xxx flags (namely, SA_PROBE,
SA_SAMPLE_RANDOM and SA_SHIRQ) from all the arch-specific headers to
linux/signal.h.  This looks like a left-over after the irq-handling code
was consolidated.  The code was moved to kernel/irq/*, but the flags are
still left per-arch.

Right now, adding a new IRQ flag to the arch-specific header, like this
patch does:
http://cvs.sourceforge.net/viewcvs.py/*checkout*/alsa/alsa-driver/utils/patches/pcsp-kernel-2.6.10-03.diff?rev=1.1
no longer works, it breaks the compilation for all other arches, unless you
add that flag to all the other arch-specific headers too.  So I think such
a clean-up makes sense.

Signed-off-by: Stas Sergeev <stsp@aknet.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] clean up kernel messages
Matt Mackall [Sun, 1 May 2005 15:59:02 +0000 (08:59 -0700)]
[PATCH] clean up kernel messages

Arrange for all kernel printks to be no-ops.  Only available if
CONFIG_EMBEDDED.

This patch saves about 375k on my laptop config and nearly 100k on minimal
configs.

Signed-off-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Exterminate PAGE_BUG
Matt Mackall [Sun, 1 May 2005 15:59:01 +0000 (08:59 -0700)]
[PATCH] Exterminate PAGE_BUG

Remove PAGE_BUG - repalce it with BUG and BUG_ON.

Signed-off-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] remove all kernel BUGs
Matt Mackall [Sun, 1 May 2005 15:59:01 +0000 (08:59 -0700)]
[PATCH] remove all kernel BUGs

This patch eliminates all kernel BUGs, trims about 35k off the typical
kernel, and makes the system slightly faster.

Signed-off-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] nice and rt-prio rlimits
Matt Mackall [Sun, 1 May 2005 15:59:00 +0000 (08:59 -0700)]
[PATCH] nice and rt-prio rlimits

Add a pair of rlimits for allowing non-root tasks to raise nice and rt
priorities. Defaults to traditional behavior. Originally written by
Chris Wright.

The patch implements a simple rlimit ceiling for the RT (and nice) priorities
a task can set.  The rlimit defaults to 0, meaning no change in behavior by
default.  A value of 50 means RT priority levels 1-50 are allowed.  A value of
100 means all 99 privilege levels from 1 to 99 are allowed.  CAP_SYS_NICE is
blanket permission.

(akpm: see http://www.uwsg.iu.edu/hypermail/linux/kernel/0503.1/1921.html for
tips on integrating this with PAM).

Signed-off-by: Matt Mackall <mpm@selenic.com>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: cio documentation
Cornelia Huck [Sun, 1 May 2005 15:59:00 +0000 (08:59 -0700)]
[PATCH] s390: cio documentation

Synchronize documentation with current interface.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: remove ioctl32 from crypto driver
Cornelia Huck [Sun, 1 May 2005 15:59:00 +0000 (08:59 -0700)]
[PATCH] s390: remove ioctl32 from crypto driver

The ioctl32_conversion routines will be deprecated: Remove them from the
crypto driver.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: remove ioctl32 from dasdcmb
Cornelia Huck [Sun, 1 May 2005 15:58:59 +0000 (08:58 -0700)]
[PATCH] s390: remove ioctl32 from dasdcmb

The ioctl32_conversion routines will be deprecated: Remove them from dasd_cmb
and handle the three cmb ioctls like all other dasd ioctls.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: don't pad cdl blocks for write requests
Horst Hummel [Sun, 1 May 2005 15:58:59 +0000 (08:58 -0700)]
[PATCH] s390: don't pad cdl blocks for write requests

The first blocks on a cdl formatted dasd device are smaller than the blocksize
of the device.  Read requests are padded with a 'e5' pattern.  Write requests
should not pad the (user) buffer with 'e5' because a write request is not
allowed to modify the buffer.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: enable write barriers in the dasd driver
Stefan Weinhuber [Sun, 1 May 2005 15:58:59 +0000 (08:58 -0700)]
[PATCH] s390: enable write barriers in the dasd driver

The DASD device driver never reorders the I/O requests and relies on the
hardware to write all data to nonvolatile storage before signaling a
successful write.  Hence, the only thing we have to do to support write
barriers is to set the queue ordered flag.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: dasd readonly attribute
Horst Hummel [Sun, 1 May 2005 15:58:59 +0000 (08:58 -0700)]
[PATCH] s390: dasd readonly attribute

The independent read-only flags in devmap, dasd_device and gendisk are not
kept in sync.  Use one bit per feature in the dasd driver and keep that bit in
sync with the gendisk bit.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: allow longer debug feature names
Michael Holzheu [Sun, 1 May 2005 15:58:58 +0000 (08:58 -0700)]
[PATCH] s390: allow longer debug feature names

The current limitation of 16 characters of the debug feature names turned out
to be insufficient.  Increase it to 64 characters.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: cmm guest sender id
Martin Schwidefsky [Sun, 1 May 2005 15:58:58 +0000 (08:58 -0700)]
[PATCH] s390: cmm guest sender id

An arbitrary guest must not be allowed to trigger cmm actions.  Only one
specific guest namely the one that serves as the resource monitor may send cmm
messages.  Add a parameter that allows to specify the guest that may send
messages.  z/VMs resource manager has the name 'VMRMSVM' which is the default.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: default storage key
Peter Oberparleiter [Sun, 1 May 2005 15:58:58 +0000 (08:58 -0700)]
[PATCH] s390: default storage key

Provide an easy way to define a non-zero storage key at compile time.  This is
useful for debugging purposes.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: fix memory holes and cleanup setup_arch
Heiko Carstens [Sun, 1 May 2005 15:58:57 +0000 (08:58 -0700)]
[PATCH] s390: fix memory holes and cleanup setup_arch

The memory setup didn't take care of memory holes and this makes the memory
management think there would be more memory available than there is in
reality.  That causes the OOM killer to kill processes even if there is enough
memory left that can be written to the swap space.

The patch fixes this by using free_area_init_node with an array of memory
holes instead of free_area_init.  Further the patch cleans up the code in
setup.c by splitting setup_arch into smaller pieces.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: idle timer setup
Martin Schwidefsky [Sun, 1 May 2005 15:58:57 +0000 (08:58 -0700)]
[PATCH] s390: idle timer setup

Fix overflow in calculation of the new tod value in stop_hz_timer and fix
wrong virtual timer list idle time in case the virtual timer is already
expired in stop_cpu_timer.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] s390: regenerate defconfig
Martin Schwidefsky [Sun, 1 May 2005 15:58:57 +0000 (08:58 -0700)]
[PATCH] s390: regenerate defconfig

Regenerate the default configuration for s390.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml ubd: handle readonly status
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:57 +0000 (08:58 -0700)]
[PATCH] uml ubd: handle readonly status

Use the set_disk_ro() API when the backing file is read-only, to mark the disk
read-only, during the ->open().  The current hack does not work when doing a
mount -o remount.

Also, mark explicitly the code paths which should no more be triggerable (I've
removed the WARN_ON(1) things).  They should actually become BUG()s probably
but I'll avoid that since I'm not so sure the change works so well.  I gave it
only some limited testing.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
CC: Jens Axboe <axboe@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: commentary about forking flag
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:56 +0000 (08:58 -0700)]
[PATCH] uml: commentary about forking flag

Add some commentary about UML internals, for a strange trick.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml - hostfs: avoid buffers
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:56 +0000 (08:58 -0700)]
[PATCH] uml - hostfs: avoid buffers

Use this:
.set_page_dirty = __set_page_dirty_nobuffers,

We already dropped the inclusion of <linux/buffer_head.h>, and we don't have a
backing block device for this FS.

"Without having looked at it, I'm sure that hostfs does not use buffer_heads.
So setting your ->set_page_dirty a_op to point at __set_page_dirty_nobuffers()
is a reasonable thing to do - it'll provide a slight speedup."

This speedup is one less spinlock held and one less conditional branch, which
isn't bad.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: redo console locking
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:56 +0000 (08:58 -0700)]
[PATCH] uml: redo console locking

Fix some console locking problems (including scheduling in atomic) and various
reorderings and cleanup in that code.  Not yet ready for 2.6.12 probably.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: fix syscall table by including $(SUBARCH)'s one, for x86-64
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:55 +0000 (08:58 -0700)]
[PATCH] uml: fix syscall table by including $(SUBARCH)'s one, for x86-64

Reuse asm-x86-64/unistd.h to build our syscall table, like x86-64 already
does.

Like for i386, we must add some #defines for all the (right!) changes UML does
to x86-64 syscall table.

Note: I noted a bogus:
[ __NR_sched_yield ] = (syscall_handler_t *) yield,

while doing this patch (which could only be a workaround for some strange bug,
but I would ignore this possibility).  I'm changing this without notice.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: quick fix syscall table for x86_64
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:55 +0000 (08:58 -0700)]
[PATCH] uml: quick fix syscall table for x86_64

Fix the moved syscall table for the x86_64 SUBARCH:

- redirect __NR_chown and such to versions aware of 32-bit UIDs,

- avoid the useless hack for sys_nfsservctl,

- use sys_sendfile64 in the table rather than sys_sendfile.

- __NR_uselib is sys_ni_syscall on x86_64 (which does not support A.OUT).

- __NR_getrlimit is sys_getrlimit, not sys_old_getrlimit

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: fix syscall table by including $(SUBARCH)'s one, for i386
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:55 +0000 (08:58 -0700)]
[PATCH] uml: fix syscall table by including $(SUBARCH)'s one, for i386

Split the i386 entry.S files into entry.S and syscall_table.S which is
included in the previous one (so actually there is no difference between them)
and use the syscall_table.S in the UML build, instead of tracking by hand the
syscall table changes (which is inherently error-prone).

We must only insert the right #defines to inject the changes we need from the
i386 syscall table (for instance some different function names); also, we
don't implement some i386 syscalls, as ioperm(), nor some TLS-related ones
(yet to provide).

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: move va_copy conditional def
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:54 +0000 (08:58 -0700)]
[PATCH] uml: move va_copy conditional def

GCC 2.95 uses __va_copy instead of va_copy.  Handle it inside compiler.h
instead of in a casual file, and avoid the risk that this breaks with a newer
compiler (which it could do).

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: inline empty proc
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:54 +0000 (08:58 -0700)]
[PATCH] uml: inline empty proc

Cleanup: make an inline of this empty proc.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: support AES i586 crypto driver
Paolo 'Blaisorblade' Giarrusso [Sun, 1 May 2005 15:58:54 +0000 (08:58 -0700)]
[PATCH] uml: support AES i586 crypto driver

We want to make possible, for the user, to enable the i586 AES implementation.
This requires a restructure.

- Add a CONFIG_UML_X86 to notify that we are building a UML for i386.

- Rename CONFIG_64_BIT to CONFIG_64BIT as is used for all other archs

- Tell crypto/Kconfig that UML_X86 is as good as X86

- Tell it that it must exclude not X86_64 but 64BIT, which will give the
  same results.

- Tell kbuild to descend down into arch/i386/crypto/ to build what's needed.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: workaround old problematic sed behaviour
Rob Landley [Sun, 1 May 2005 15:58:54 +0000 (08:58 -0700)]
[PATCH] uml: workaround old problematic sed behaviour

Old versions of sed from 1998 (predating the first release of gcc 2.95, but
still in use by debian stable) don't understand the single-line version of the
sed append command.  Since newer versions of sed still understand the...
ahem, "vintage" form of the command, change our code to use that.

Signed-off-by: Rob Landley <rob@landley.net>
Acked-by: Ian McDonald <imcdnzl@gmail.com>
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: fix handling of no fpx_regs
Andree Leidenfrost [Sun, 1 May 2005 15:58:53 +0000 (08:58 -0700)]
[PATCH] uml: fix handling of no fpx_regs

      Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>

Fix the error path, which is triggered when the processor misses the fpx
regs (i.e.  the "fxsr" cpuinfo feature).  For instance by VIA C3 Samuel2.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: add nfsd syscall when nfsd is modular
blaisorblade@yahoo.it [Sun, 1 May 2005 15:58:53 +0000 (08:58 -0700)]
[PATCH] uml: add nfsd syscall when nfsd is modular

This trick is useless, because sys_ni.c will handle this problem by itself,
like it does even on UML for other syscalls.

Also, it does not provide the NFSD syscall when NFSD is compiled as a
module, which is a big problem.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] uml: fix oops related to exception table
Jeff Dike [Sun, 1 May 2005 15:58:53 +0000 (08:58 -0700)]
[PATCH] uml: fix oops related to exception table

      Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>

Prevent the kernel from oopsing during the extable sorting, as it can do
now, because the extable is in the readonly section of the binary.

Jeff says: The exception table turned RO in 2.6.11-rc3-mm1 for some reason.
Moving it causes it to land in the writable data section of the binary.

Paolo says: This patch fixes a oops on startup, which can be easily
triggered by compiling with CONFIG_MODE_TT disabled, and STATIC_LINK either
disabled or enabled.  The resulting kernel will always Oops on startup,
after printing this simple output:

I've verified, by binary search on the BitKeeper repository (synced up as
of 2.6.12-rc2), starting from the range 2.6.11-2.6.12-rc1, that this bug
shows up on BitKeeper revisions in the range [@1.1994.11.168,+inf), i.e.
starting from this:

[PATCH] lib/sort: Replace insertion sort in exception tables

Since UML does not use the exception table, it's likely that insertion sort
didn't happen to write anything on the table.

Signed-off-by: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] hp100: fix card names
Pavel Machek [Sun, 1 May 2005 15:58:52 +0000 (08:58 -0700)]
[PATCH] hp100: fix card names

Those cards really need A in their names. Otherwise it is pretty hard
to find anything about them on the net.

Signed-off-by: Pavel Machek <pavel@suse.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] x86_64: saved_command_line overflow fix
Alexander Nyberg [Sun, 1 May 2005 15:58:52 +0000 (08:58 -0700)]
[PATCH] x86_64: saved_command_line overflow fix

This strcpy can run off the end of saved_command_line, and we don't need it any more anyway.

Signed-off-by: Alexander Nyberg <alexn@telia.com>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] x86-64: Handle empty E820 regions correctly
Venkatesh Pallipadi [Sun, 1 May 2005 15:58:52 +0000 (08:58 -0700)]
[PATCH] x86-64: Handle empty E820 regions correctly

Brings sanitize_e820_map() in x86-64 in sync with that of i386.

x86_64 version was missing the changes from this patch.
http://linux.bkbits.net:8080/linux-2.6/cset@3e5e4083Y3HevldZl5KCy94V4DcZww?nav=index.html|src/|src/arch|src/arch/i386|src/arch/i386/kernel|related/arch/i386/kernel/setup.c

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Linux 2.6.x VM86 interrupt emulation fixes
Pavel Pisa [Sun, 1 May 2005 15:58:52 +0000 (08:58 -0700)]
[PATCH] Linux 2.6.x VM86 interrupt emulation fixes

Patch solves VM86 interrupt emulation deadlock on SMP systems.  The VM86
interrupt emulation has been heavily tested and works well on UP systems
after last update, but it seems to deadlock when we have used it on SMP/HT
boxes now.

It seems, that disable_irq() cannot be called from interrupts, because it
waits until disabled interrupt handler finishes
(/kernel/irq/manage.c:synchronize_irq():while(IRQ_INPROGRESS);).  This
blocks one CPU after another.  Solved by use disable_irq_nosync.

There is the second problem.  If IRQ source is fast, it is possible, that
interrupt is sometimes processed and re-enabled by the second CPU, before
it is disabled by the first one, but negative IRQ disable depths are not
allowed.  The spinlocking and disabling IRQs over call to
disable_irq_nosync/enable_irq is the only solution found reliable till now.

Signed-off-by: Michal Sojka <sojkam1@control.felk.cvut.cz>
Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Increase number of e820 entries hard limit from 32 to 128
Venkatesh Pallipadi [Sun, 1 May 2005 15:58:51 +0000 (08:58 -0700)]
[PATCH] Increase number of e820 entries hard limit from 32 to 128

The specifications that talk about E820 map doesn't have an upper limit on
the number of e820 entries.  But, today's kernel has a hard limit of 32.
With increase in memory size, we are seeing the number of E820 entries
reaching close to 32.  Patch below bumps the number upto 128.

The patch changes the location of EDDBUF in zero-page (as it comes after E820).
As, EDDBUF is not used by boot loaders, this patch should not have any effect
on bootloader-setup code interface.

Patch covers both i386 and x86-64.

Tested on:
* grub booting bzImage
* lilo booting bzImage with EDID info enabled
* pxeboot of bzImage

Side-effect:
bss increases by ~ 2K and init.data increases by ~7.5K
on all systems, due to increase in size of static arrays.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] x86_64: interrupt handling fix
Andi Kleen [Sun, 1 May 2005 15:58:51 +0000 (08:58 -0700)]
[PATCH] x86_64: interrupt handling fix

- Initialize workmask correctly on interrupt signal handling

- Readd missing cli's in the interrupt return path.

Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] cpuid x87 bit on AMD falsely marked as PNI
Zwane Mwaikambo [Sun, 1 May 2005 15:58:51 +0000 (08:58 -0700)]
[PATCH] cpuid x87 bit on AMD falsely marked as PNI

http://bugme.osdl.org/show_bug.cgi?id=4426

vendor_id       : AuthenticAMD
cpu family      : 6
model           : 10
model name      : AMD Athlon(tm) XP
stepping        : 0
cpu MHz         : 2204.807
<snipped>
cpuid level     : 1
wp              : yes
flags           : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov
pat pse36 mmx fxsr sse pni syscall mmxext 3dnowext 3dnow
bogomips        : 4358.14

We're marking bit 0 of extended function 0x80000001 cpuid as PNI support on
AMD processors, when it actually denotes x87 FPU present.  Patch for i386
and x86_64 below.

Signed-off-by: Zwane Mwaikambo <zwane@arm.linux.org.uk>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] hda_intel: Intel ESB2 support
Jason Gaston [Sun, 1 May 2005 15:58:50 +0000 (08:58 -0700)]
[PATCH] hda_intel: Intel ESB2 support

This adds the Intel ESB2 HD Audio DID to the hda_intel.c audio driver.

Signed-off-by:  Jason Gaston <Jason.d.gaston@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] irq and pci_ids for Intel ICH7DH & ICH7-M DH
Jason Gaston [Sun, 1 May 2005 15:58:50 +0000 (08:58 -0700)]
[PATCH] irq and pci_ids for Intel ICH7DH & ICH7-M DH

This patch adds the Intel ICH7DH and ICH7-M DH DID's to the irq.c and
pci_ids.h files.

Signed-off-by:  Jason Gaston <Jason.d.gaston@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] i386: fix hpet for systems that don't support legacy replacement
john stultz [Sun, 1 May 2005 15:58:50 +0000 (08:58 -0700)]
[PATCH] i386: fix hpet for systems that don't support legacy replacement

Currently the i386 HPET code assumes the entire HPET implementation from
the spec is present.  This breaks on boxes that do not implement the
optional legacy timer replacement functionality portion of the spec.

This patch, which is very similar to my x86-64 patch for the same issue,
fixes the problem allowing i386 systems that cannot use the HPET for the
timer interrupt and RTC to still use the HPET as a time source.  I've
tested this patch on a system systems without HPET, with HPET but without
legacy timer replacement, as well as HPET with legacy timer replacement.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] CPUID bug and inconsistency fix
H. Peter Anvin [Sun, 1 May 2005 15:58:49 +0000 (08:58 -0700)]
[PATCH] CPUID bug and inconsistency fix

The recent support for K8 multicore was misported from x86-64 to i386, due
to an unnecessary inconsistency between the CPUID code.  Sure, there is are
no x86-64 VIA chips yet, but it should happen eventually.

This patch fixes the i386 bug as well as makes x86-64 match i386 in the
handing of the CPUID array.

Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Enable write combining for server works LE rev > 6
Lee Revell [Sun, 1 May 2005 15:58:49 +0000 (08:58 -0700)]
[PATCH] Enable write combining for server works LE rev > 6

Enable write combining for server works LE rev > 6 per
http://www.ussg.iu.edu/hypermail/linux/kernel/0104.3/1007.html

Signed-Off-By: Lee Revell <rlrevell@joe-job.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] x86: entry.S trap return fixes
Stas Sergeev [Sun, 1 May 2005 15:58:49 +0000 (08:58 -0700)]
[PATCH] x86: entry.S trap return fixes

do_debug() and do_int3() return void.

This patch fixes the CONFIG_KPROBES variant of do_int3() to return void too
and adjusts entry.S accordingly.

Signed-off-by: Stas Sergeev <stsp@aknet.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] x86 reboot: Add reboot fixup for gx1/cs5530a
Jaya Kumar [Sun, 1 May 2005 15:58:49 +0000 (08:58 -0700)]
[PATCH] x86 reboot: Add reboot fixup for gx1/cs5530a

This patch by Jaya Kumar introduces a generic infrastructure to deal with
x86 chipsets with nonstandard reset sequences, and adds support for the
Geode gx1/cs5530a chipset.

Signed-off-by: Jaya Kumar <jayalk@intworks.biz>
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] check nmi watchdog is broken
Jack F Vogel [Sun, 1 May 2005 15:58:48 +0000 (08:58 -0700)]
[PATCH] check nmi watchdog is broken

A bug against an xSeries system showed up recently noting that the
check_nmi_watchdog() test was failing.

I have been investigating it and discovered in both i386 and x86_64 the
recent change to the routine to use the cpu_callin_map has uncovered a
problem.  Prior to that change, on an SMP box, the test was trivally
passing because all cpu's were found to not yet be online, but now with the
callin_map they are discovered, it goes on to test the counter and they
have not yet begun to increment, so it announces a CPU is stuck and bails
out.

On all the systems I have access to test, the announcement of failure is
also bougs...  by the time you can login and check /proc/interrupts, the
NMI count is happily incrementing on all CPUs.  Its just that the test is
being done too early.

I have tried moving the call to the test around a bit, and it was always
too early.  I finally hit on this proposed solution, it delays the routine
via a late_initcall(), seems like the right solution to me.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Cc: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] i386/x86_64 segment register access update
H. J. Lu [Sun, 1 May 2005 15:58:48 +0000 (08:58 -0700)]
[PATCH] i386/x86_64 segment register access update

The new i386/x86_64 assemblers no longer accept instructions for moving
between a segment register and a 32bit memory location, i.e.,

        movl (%eax),%ds
        movl %ds,(%eax)

To generate instructions for moving between a segment register and a
16bit memory location without the 16bit operand size prefix, 0x66,

        mov (%eax),%ds
        mov %ds,(%eax)

should be used. It will work with both new and old assemblers. The
assembler starting from 2.16.90.0.1 will also support

        movw (%eax),%ds
        movw %ds,(%eax)

without the 0x66 prefix. I am enclosing patches for 2.4 and 2.6 kernels
here. The resulting kernel binaries should be unchanged as before, with
old and new assemblers, if gcc never generates memory access for

               unsigned gsindex;
               asm volatile("movl %%gs,%0" : "=g" (gsindex));

If gcc does generate memory access for the code above, the upper bits
in gsindex are undefined and the new assembler doesn't allow it.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] fix i386 memcpy
Denis Vlasenko [Sun, 1 May 2005 15:58:48 +0000 (08:58 -0700)]
[PATCH] fix i386 memcpy

This patch shortens non-constant memcpy() by two bytes and fixes spurious
out-of-line constant memcpy().

# size vmlinux.org vmlinux
   text    data     bss     dec     hex filename
3954591 1553426  236544 5744561  57a7b1 vmlinux.org
3952615 1553426  236544 5742585  579ff9 vmlinux

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: reverse prediction on spinlock busy loop code
Jake Moilanen [Sun, 1 May 2005 15:58:47 +0000 (08:58 -0700)]
[PATCH] ppc64: reverse prediction on spinlock busy loop code

On our raw spinlocks, we currently have an attempt at the lock, and if we do
not get it we enter a spin loop.  This spinloop will likely continue for
awhile, and we pridict likely.

Shouldn't we predict that we will get out of the loop so our next instructions
are already prefetched.  Even when we miss because the lock is still held, it
won't matter since we are waiting anyways.

I did a couple quick benchmarks, but the results are inconclusive.

16-way 690 running specjbb with original code
# ./specjbb 3000 16 1 1 19 30 120
    ...
Valid run, Score is 59282

16-way 690 running specjbb with unlikely code
# ./specjbb 3000 16 1 1 19 30 120
    ...
Valid run, Score is 59541

I saw a smaller increase on a JS20 (~1.6%)

JS20 specjbb w/ original code
# ./specjbb 400 2 1 1 19 30 120
   ...
Valid run, Score is 20460

JS20 specjbb w/ unlikely code
# ./specjbb 400 2 1 1 19 30 120
   ...
Valid run, Score is 20803

Anton said:

Mispredicting the spinlock busy loop also means we slow down the rate at which
we do the loads which can be good for heavily contended locks.

Note: There are some gcc issues with our default build and branch prediction,
but a CONFIG_POWER4_ONLY build should emit them correctly.  I'm working with
Alan Modra on it now.

Signed-off-by: Jake Moilanen <moilanen@austin.ibm.com>
Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] use smp_mb/wmb/rmb where possible
akpm@osdl.org [Sun, 1 May 2005 15:58:47 +0000 (08:58 -0700)]
[PATCH] use smp_mb/wmb/rmb where possible

Replace a number of memory barriers with smp_ variants.  This means we won't
take the unnecessary hit on UP machines.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: use smp_mb and smp_wmb
Anton Blanchard [Sun, 1 May 2005 15:58:47 +0000 (08:58 -0700)]
[PATCH] ppc64: use smp_mb and smp_wmb

Use smp_mb and smp_wmb. In particular smp_wmb is lighter weight than wmb.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: enforce medium thread priority in hypervisor calls
Anton Blanchard [Sun, 1 May 2005 15:58:46 +0000 (08:58 -0700)]
[PATCH] ppc64: enforce medium thread priority in hypervisor calls

Calls into the hypervisor do not raise the thread priority.  Ensure we are
running at medium priority upon entry to the hypervisor.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: firmware workaround
Anton Blanchard [Sun, 1 May 2005 15:58:46 +0000 (08:58 -0700)]
[PATCH] ppc64: firmware workaround

Recent gcc 4.0 testing uncovered a firmware issue.  Some properties are larger
than 31 bytes and due to gcc 4.0s better stack allocation this overflow ran
over non volatile register storage.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: remove unnecessary include
Anton Blanchard [Sun, 1 May 2005 15:58:46 +0000 (08:58 -0700)]
[PATCH] ppc64: remove unnecessary include

We no longer use any ppcdebug stuff in a.out.h, so remove the define.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: noexec fixes
Anton Blanchard [Sun, 1 May 2005 15:58:45 +0000 (08:58 -0700)]
[PATCH] ppc64: noexec fixes

There were a few issues with the ppc64 noexec support:

The 64bit ABI has a non executable stack by default.  At the moment 64bit apps
require a PT_GNU_STACK section in order to have a non executable stack.

Disable the read implies exec workaround on the 64bit ABI.  The 64bit
toolchain has never had problems with incorrect mmap permissions (the 32bit
has, thats why we need to retain the workaround).

With these fixes as well as a gcc fix from Alan Modra (that was recently
committed) 64bit apps work as expected.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] PPC64: Remove hot busy-wait loop in __hash_page
Olof Johansson [Sun, 1 May 2005 15:58:45 +0000 (08:58 -0700)]
[PATCH] PPC64: Remove hot busy-wait loop in __hash_page

It turns out that our current __hash_page code will do a very hot busy-wait
loop waiting on _PAGE_BUSY to be cleared.  It even does ldarx/stdcx in the
loop, which will bounce reservations around like crazy if there's more than
one CPU spinning on the same PTE (or even another PTE in the same
reservation granule).  The end result is that each fault takes longer when
there's contention, which in turn increases the chance of another thread
hitting the same fault and also piling up.  Not pretty.

There's two options here:
1. Do an out-of-line busy loop a'la spinlocks with just loads (no
   reserves)
2. Just bail and refault if needed.

(2) makes sense here: If the PTE is busy, chances are it's in flux anyway
and the other code path making a change might just be ready to hash it.

This fixes a stampede seen on a large-ish system where a multithreaded
HPC app faults in the same text pages on several cpus at the same time.

Signed-off-by: Olof Johansson <olof@lixom.net>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: tell firmware about kernel capabilities
Paul Mackerras [Sun, 1 May 2005 15:58:45 +0000 (08:58 -0700)]
[PATCH] ppc64: tell firmware about kernel capabilities

On pSeries systems, according to the platform architecture specs, we are
supposed to be supplying a structure to firmware that tells firmware about
our capabilities, such as which version of the data structures that
describe available memory we are expecting to see.  The way we end up
having to supply this data structure is a bit gross, since it was designed
for AIX and doesn't suit us very well.  This patch adds the code to supply
this data structure to the firmware.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: update to use the new 4L headers
Benjamin Herrenschmidt [Sun, 1 May 2005 15:58:44 +0000 (08:58 -0700)]
[PATCH] ppc64: update to use the new 4L headers

This patch converts ppc64 to use the generic pgtable-nopud.h instead of the
"fixup" header.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: nvram cleanups
akpm@osdl.org [Sun, 1 May 2005 15:58:44 +0000 (08:58 -0700)]
[PATCH] ppc64: nvram cleanups

- Fix

  arch/ppc64/kernel/nvram.c:342: warning: `part' might be used uninitialized in this function

- Various codingstyle tweaks.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: Fix irq parsing on powermac
Paul Mackerras [Sun, 1 May 2005 15:58:44 +0000 (08:58 -0700)]
[PATCH] ppc64: Fix irq parsing on powermac

When I tried Ben's patches to the powermac sound driver on my G5, I found
that it was taking enormous numbers of sound DMA transmit interrupts.  This
turned out to be because it was incorrectly configured as level-sensitive
instead of edge-sensitive, which in turn was because the code that parses
the interrupt tree that Open Firmware gives us was incorrectly assigning
another device the same irq number as the sound DMA transmit interrupt
(i.e.  1).

This patch fixes the problem, in a somewhat quick and dirty way for now,
but one which will work for all the machines we currently run on.
Ultimately Ben and I want to do something more general and robust, but this
should go in for 2.6.12.

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ppc64: remove unused argument to create_slbe
Olof Johansson [Sun, 1 May 2005 15:58:44 +0000 (08:58 -0700)]
[PATCH] ppc64: remove unused argument to create_slbe

Remove vsid argument to create_slbe, since it's no longer used.

Spotted by R Sharada.

Signed-off-by: Olof Johansson <olof@austin.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>