]> git.hungrycats.org Git - linux/log
linux
22 years ago[PATCH] fbcon: new scrolling mode: YPAN + REDRAW
Antonino Daplas [Tue, 6 Jul 2004 16:19:02 +0000 (09:19 -0700)]
[PATCH] fbcon: new scrolling mode: YPAN + REDRAW

Added a new scrolling mode (SCROLL_PAN_REDRAW) to fbcon that should greatly
benefit unaccelerated drivers such as VESA fbdev.  An increase of 3-10 times
in scrolling speed can be expected.

Currently, fbcon has 4 different scrolling methods (1-4).  Potentially, we can
have 6.  This patch implements SCROLL_PAN_REDRAW (5).  SCROLL_WRAP_REDRAW (6)
is still unimplemented.

Scroll Mode Operation YPan YWrap
+++++++++++++++++++++++++++++++++++++++++++++++++++++
1. SCROLL_ACCEL copyarea No No
2. SCROLL_REDRAW imageblit No No
3. SCROLL_PAN copyarea Yes No
4. SCROLL_WRAP copyarea No Yes
5. SCROLL_PAN_REDRAW  imageblit Yes No
6. SCROLL_WRAP_REDRAW imageblit No Yes
+++++++++++++++++++++++++++++++++++++++++++++++++++++
Note 1:  I've changed the nomenclature to increase clarity:

SCROLL_ACCEL = SCROLL_MOVE
SCROLL_REDRAW = SCROLL_REDRAW
SCROLL_PAN = SCROLL_PAN_MOVE
SCROLL_WRAP = SCROLL_WRAP_MOVE

To demonstrate the effect of each of the scrolling methods on
an unaccelerated PCI/AGP-based driver (vesafb), I used a simple
benchmark (time cat linux/MAINTAINERS - a 50K text file). The
framebuffer is set at:

1024x768-8bpp, 8x16 font, yres_virtual = 2*yres

1. SCROLL_MOVE:
real    5m50.277s
user    0m0.001s
sys     5m50.227s

Almost 6 minutes for a 50K text file.  Using soft copyarea on a
PCI-based card is just too slow (because copyarea has to
read from the framebuffer memory).

2. SCROLL_PAN_MOVE
scrollmode: SCROLL_PAN
real    0m8.592s
user    0m0.000s
sys     0m8.586s

Using ypan with copyarea dramatically improves the scrolling.  However, the
scrolling action is jerky (fast during the panning stages, slows down during
the copyarea stages).

3. SCROLL_REDRAW (this is the default scrolling mode)
real    0m3.189s
user    0m0.000s
sys     0m3.170s

Simply eliminating the copyarea, even without ypanning, makes it faster than
SCROLL_PAN_MOvE.  Plus, the scrolling action is smoother.

So, if we combine YPanning with imageblit (PAN_REDRAW), we get this:

4. SCROLL_PAN_REDRAW
real    0m0.520s
user    0m0.000s
sys     0m0.518s

That's almost 6x faster than SCROLL_REDRAW.  Increasing the amount of video
RAM still increases the speed, but not very dramatically.  Higher than 16 MB,
the increase is negligible.

Using an accelerated driver, we see almost the same effect but not as
dramatically:

1. SCROLL_MOVE - accel
real    0m3.112s
user    0m0.000s
sys     0m3.112s

2. SCROLL_REDRAW - accel
real    0m2.604s
user    0m0.000s
sys     0m2.603s

Redraw is still faster than move, but not much.

3. SCROLL_PAN_MOVE - accel
real    0m0.203s
user    0m0.000s
sys     0m0.202s

4. SCROLL_PAN_REDRAW - accel
real    0m0.326s
user    0m0.002s
sys     0m0.323s

This is one exception.  If panning is enabled, move is actually faster than
redraw.  As to why, I don't know.

So based on the above, fbcon will choose the scrolling method based on the
following preference:

Ypan/Ywrap > accel imageblit > accel copyarea >
soft imageblit > (soft copyarea)

Note 2: Exception: accel copyarea > accel imageblit if Pan/Wrap is enabled.

Note 3: soft copyarea will be avoided by fbcon as much as possible unless
there is a specific override, ie., FBINFO_READS_FAST flag is set.  If this
flag is set, fbcon will prefer copyarea over imageblit, accel or soft.

As a final note, in order for fbcon to use the best scrolling method, the
low-level drivers must provide the correct hinting by setting the
FBINFO_HWACCEL_* flags.

To vesafb users: boot vesafb like this for fastest scrolling action:

video=vesafb:ypan,vram:16

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] fbcon: refinements for fbcon
Antonino Daplas [Tue, 6 Jul 2004 16:18:51 +0000 (09:18 -0700)]
[PATCH] fbcon: refinements for fbcon

1.  From Petr Vandrovec <vandrove@vc.cvut.cz>.

Begin quote

        yesterday I finally found why 2.6.x has problems with taking
        over vgacon.  Problem is that take_over_console does:

   if (IS_VISIBLE)
     save_screen(i);
   ...
   visual_init(i, 0);
   ...

        so code nicely saves screen contents, and calls visual_init,
        which calls fbcon's init, which in turn can issue vc_resize.  And this
        vc_resize will trigger read from vc->vc_origin.  But vc_origin still
        points to videomemory, not to the buffer where save_screen() copied
        its data.

End quote

So updating vc->vc_origin to point to screenbuf after the previous console
deinitializes allows fbcon to correctly update the screen from the screen
buffer.  This change makes the clearing of vram (using fillrect) unnecessary.

2.  If FBINFO_MISC_MODESWITCH flag is set in info, do the set_par() in
fbcon_switch() instead of piggy-backing the set_par() on fbcon_resize().  This
preserves the current var, skipping the unnecessary fb_find_mode() step.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] fbcon mode switching fix
Antonino Daplas [Tue, 6 Jul 2004 16:18:40 +0000 (09:18 -0700)]
[PATCH] fbcon mode switching fix

Ugly workaround.  When switching from KD_GRAPHICS to KD_TEXT, the event is
captured at fbcon_blank() allowing fbcon to reinitialize the hardware.
However, some hardware requires the reinitialization to be done
immediately, others require it to be done later.  Others may need it to be
done immediately and later, this is the worst case.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] s390: cpu hotplug support
Martin Schwidefsky [Tue, 6 Jul 2004 16:18:28 +0000 (09:18 -0700)]
[PATCH] s390: cpu hotplug support

From: Ursula Braun-Krahl <braunu@de.ibm.com>
From: Heiko Carstens <heiko.carstens@de.ibm.com>
From: Gerald Schaefer <geraldsc@de.ibm.com>
From: Martin Schwidefsky <schwidefsky@de.ibm.com>

Add s390 architecture support for cpu hotplug.

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>
22 years ago[PATCH] s390: cpu-idle notifier
Martin Schwidefsky [Tue, 6 Jul 2004 16:18:17 +0000 (09:18 -0700)]
[PATCH] s390: cpu-idle notifier

This patch introduces a notifier chain for cpu idle.  There are two events
CPU_IDLE and CPU_NOT_IDLE that are called just before the cpu goes to sleep
and right after the cpu woke up again.  The notifier is used to simplify the
no-hz-timer-in-idle feature and the virtual cpu timers.  In addition the
virtual cpu timer functions have been moved to arch/s390/kernel/vtime.c, which
gives a nice separation between the timer functions related to real time and
the timer functions related to virtual cpu time.

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>
22 years ago[PATCH] s390: ctc driver changes
Martin Schwidefsky [Tue, 6 Jul 2004 16:18:05 +0000 (09:18 -0700)]
[PATCH] s390: ctc driver changes

From: Peter Tiedemann <ptiedem@de.ibm.com>

ctc driver changes:
 - Make use of the debug feature to ease debugging.
 - ctctty: use dev_alloc_name to allocate a network device name.
 - ctctty: avoid deadlock of ctc_tty_close vs ctc_tty_flush_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>
22 years ago[PATCH] s390: common i/o layer
Martin Schwidefsky [Tue, 6 Jul 2004 16:17:53 +0000 (09:17 -0700)]
[PATCH] s390: common i/o layer

From: Arnd Bergmann <arndb@de.ibm.com>

Common i/o layer changes:
 - Reorder checking and setting of the ccw device id.

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>
22 years ago[PATCH] s390: dasd driver changes
Martin Schwidefsky [Tue, 6 Jul 2004 16:17:42 +0000 (09:17 -0700)]
[PATCH] s390: dasd driver changes

From: Horst Hummel <horst.hummel@de.ibm.com>

dasd device driver changes:
 - Set and check the retry counter in start_IO for all requests.

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>
22 years ago[XFRM]: Convert XFRM_MSG_* macros to an enum.
Herbert Xu [Mon, 5 Jul 2004 09:42:08 +0000 (02:42 -0700)]
[XFRM]: Convert XFRM_MSG_* macros to an enum.

This way we will not forget to update XFRM_MSG_MAX in
the future.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[NET]: Save space for dst underflow message.
Hideaki Yoshifuji [Mon, 5 Jul 2004 09:38:11 +0000 (02:38 -0700)]
[NET]: Save space for dst underflow message.

Save space for 'dst underflow' message.

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[TCP]: Inline message.
Hideaki Yoshifuji [Mon, 5 Jul 2004 09:36:42 +0000 (02:36 -0700)]
[TCP]: Inline message.

D: [TCP] Diet using unknown

22 years ago[IPV6]: Fix flags for ndisc dst.
Hideaki Yoshifuji [Mon, 5 Jul 2004 09:35:28 +0000 (02:35 -0700)]
[IPV6]: Fix flags for ndisc dst.

Because RTF_LOCAL is for local unicast address,
it was wrong to set RTF_LOCAL to ndisc dst.
This patch also adds some comment on other fields.

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[VLAN]: Do not access released memory.
Petr Vandrovec [Mon, 5 Jul 2004 09:33:33 +0000 (02:33 -0700)]
[VLAN]: Do not access released memory.

22 years agoMerge http://linux-mh.bkbits.net/bluetooth-2.6
David S. Miller [Mon, 5 Jul 2004 09:27:41 +0000 (02:27 -0700)]
Merge http://linux-mh.bkbits.net/bluetooth-2.6
into nuts.davemloft.net:/disk1/BK/net-2.6

22 years ago[IPV6] Bring lo up before setting other interface up.
Hideaki Yoshifuji [Mon, 5 Jul 2004 09:26:34 +0000 (02:26 -0700)]
[IPV6] Bring lo up before setting other interface up.

IPv6 was not configured appropriately without lo.
Noticed by / tested by Russell King <rmk+lkml@arm.linux.org.uk>.

Signed-off-by: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[AH4]: Harmonization of output function.
Herbert Xu [Mon, 5 Jul 2004 09:02:18 +0000 (02:02 -0700)]
[AH4]: Harmonization of output function.

This is another step towards the union of the tunnel mode encapsulation
between transforms.  As there are significant differences between the
tunnel encapsulation of IPv4 and IPv6, I'll be dealing with IPv4 only
for now.

This particular patch rearranges the code in ah_output to isolate the
tunnel mode encapsulation.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[AX25]: ax25_rt_add() locking fix.
Andrew Morton [Mon, 5 Jul 2004 09:00:10 +0000 (02:00 -0700)]
[AX25]: ax25_rt_add() locking fix.

Fix deadlock identified by the Stanford locking checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[LAPB]: lapb_unregister() locking fix.
Andrew Morton [Mon, 5 Jul 2004 08:59:31 +0000 (01:59 -0700)]
[LAPB]: lapb_unregister() locking fix.

Fix deadlock identified by the Stanford locking checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[AX25]: ax25_ds_idletimer_expiry() locking fix.
Andrew Morton [Mon, 5 Jul 2004 08:58:48 +0000 (01:58 -0700)]
[AX25]: ax25_ds_idletimer_expiry() locking fix.

Fix deadlock identified by the Stanford locking checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[ROSE]: rose_route locking fix
Andrew Morton [Mon, 5 Jul 2004 08:57:53 +0000 (01:57 -0700)]
[ROSE]: rose_route locking fix

Fix deadlock in rose_del_loopback_node().  Found by the Stanford locking
checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[ESP4]: Merge NAT-T code in esp_output().
Herbert Xu [Mon, 5 Jul 2004 08:54:24 +0000 (01:54 -0700)]
[ESP4]: Merge NAT-T code in esp_output().

22 years ago[NET]: Allow large MTU on dummy net device.
Stephen Hemminger [Mon, 5 Jul 2004 08:43:44 +0000 (01:43 -0700)]
[NET]: Allow large MTU on dummy net device.

It is useful for testing to allow larger MTU value to be set on
the dummy network device. The current code limits it to valid Ether mtu's.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[NETLINK]: Return err in netlink_connect.
Herbert Xu [Mon, 5 Jul 2004 08:42:39 +0000 (01:42 -0700)]
[NETLINK]: Return err in netlink_connect.

This patch makes netlink_connect() return the value of err instead of 0.
It doesn't actually make any difference since the current implementation
of netlink_autobind() never fails.  But since we went to all this trouble
to check the return status of autobind, might as well return the correct
value :)

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[NETLINK]: Check connect address.
Herbert Xu [Mon, 5 Jul 2004 08:41:28 +0000 (01:41 -0700)]
[NETLINK]: Check connect address.

22 years ago[IPV4]: Calculate default broadcast even when using SIOCSIGNETMASK.
Jaap Keuter [Mon, 5 Jul 2004 08:39:17 +0000 (01:39 -0700)]
[IPV4]: Calculate default broadcast even when using SIOCSIGNETMASK.

While getting hands-on with netkit (www.netkit.org), a networking
simulation environment based on UML, it struck me that ifconfig wasn't
capable of calculating the proper broadcast address for a subnetted
interface. Some browsing through newsgroups and on the Debian package
site (nettools), showed that this leads to misconfigured interfaces and
a couple of bugreports on ifconfig.

Digging a bit deeper revealed that it actually is an issue with
SIOIFNETMASK. Once you bring up an interface with SIOIFADDR, a classfull
netmask and broadcast address is set (if applicable for the type of
interface), in order to get a properly configured interface. But if you
subnet the network using SIOIFNETMASK no proper broadcast address is
set. So you always have to calculate it yourself, obviously leading to
configuration errors.

This patch takes care of this. First of all it doesn't change
existing functionality, eg. a command like 'ifconfig eth0 192.168.1.1
netmask 255.255.255.240 broadcast 192.168.1.0' still works. But if you
omit the broadcast address, a proper 'all ones' broadcast address for the
subnet is set. 'ifconfig eth0 192.168.1.1 netmask 255.255.255.240' gives
you 'eth0 inet addr:192.168.1.1 Bcast:192.168.1.15 Mask:255.255.255.240'
and this should solve some real life problems.

Signed-off-by: Jaap Keuter <jaap.keuter@xs4all.nl>
Signed-off-by: David S. Miller <davem@redhat.com>
22 years ago[PKT_SCHED]: Rip out requeue stat addition, user ABI breaker.
David S. Miller [Mon, 5 Jul 2004 08:28:18 +0000 (01:28 -0700)]
[PKT_SCHED]: Rip out requeue stat addition, user ABI breaker.

22 years ago[PKT_SCHED]: Pass NET_XMIT_* status properly back through classifiers.
Jamal Hadi Salim [Mon, 5 Jul 2004 08:25:34 +0000 (01:25 -0700)]
[PKT_SCHED]: Pass NET_XMIT_* status properly back through classifiers.

22 years agoMerge bk://drm.bkbits.net/drm-2.6
Linus Torvalds [Mon, 5 Jul 2004 03:54:07 +0000 (20:54 -0700)]
Merge bk://drm.bkbits.net/drm-2.6
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years agowhitespace cleanup in radeon.h
Dave Airlie [Tue, 6 Jul 2004 08:09:38 +0000 (18:09 +1000)]
whitespace cleanup in radeon.h

22 years agotypo in radeon_state.c
Dave Airlie [Tue, 6 Jul 2004 07:37:55 +0000 (17:37 +1000)]
typo in radeon_state.c

22 years agotypo..
Dave Airlie [Tue, 6 Jul 2004 07:19:40 +0000 (17:19 +1000)]
typo..

22 years ago[PATCH] kyrofb warning fix
Andrew Morton [Mon, 5 Jul 2004 03:27:16 +0000 (20:27 -0700)]
[PATCH] kyrofb warning fix

drivers/video/kyro/fbdev.c: In function `kyrofb_probe':
drivers/video/kyro/fbdev.c:736: warning: passing arg 1 of `memset' makes pointer from integer without a cast

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] remove allowdma0 documentation
Adrian Bunk [Mon, 5 Jul 2004 03:27:05 +0000 (20:27 -0700)]
[PATCH] remove allowdma0 documentation

allowdma0 is gone in 2.6, the patch below removes the mentionings of this
option in the documentation.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] spurious remap_file_pages() -EINVAL
William Lee Irwin III [Mon, 5 Jul 2004 03:26:54 +0000 (20:26 -0700)]
[PATCH] spurious remap_file_pages() -EINVAL

As ->vm_private_data is used as a cursor for swapout of VM_NONLINEAR vmas,
the check for NULL ->vm_private_data or VM_RESERVED is too strict, and
should allow VM_NONLINEAR vmas with non-NULL ->vm_private_data.

This fixes an issue on 2.6.7-mm5 where system calls to remap_file_pages()
spuriously failed while under memory pressure.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] __bdevname leak fix
Christoph Hellwig [Mon, 5 Jul 2004 03:26:43 +0000 (20:26 -0700)]
[PATCH] __bdevname leak fix

The __bdevname library function is leaking module references due to

__bdevname
  ->get_gendisk
    ->kobj_lookup
      ->ata_probe
        ->get_disk
          ->try_module_get

What we're trying to do in there is too ambitious.  Change it to just print
the major and minor.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] gcc 3.5 fixes #2
Anton Blanchard [Mon, 5 Jul 2004 03:26:31 +0000 (20:26 -0700)]
[PATCH] gcc 3.5 fixes #2

gcc 3.5 is warning about unused static variables, add __attribute_unused__
to the 2 places to silence it.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] gcc 3.5 fixes
Anton Blanchard [Mon, 5 Jul 2004 03:26:20 +0000 (20:26 -0700)]
[PATCH] gcc 3.5 fixes

gcc 3.5 is warning about static vs non static function declarations.  The
following patch removes function prototypes in .h files where possible and
changes prototypes to be static elsewhere.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] force O_LARGEFILE in sys_swapon() and sys_swapoff()
William Lee Irwin III [Mon, 5 Jul 2004 03:26:08 +0000 (20:26 -0700)]
[PATCH] force O_LARGEFILE in sys_swapon() and sys_swapoff()

For 32-bit, one quickly discovers that swapon() is not given an fd already
opened with O_LARGEFILE to act upon and the forcing of O_LARGEFILE for
64-bit is irrelevant, as the system call's argument is a path.  So this
patch manually forces it for swapon() and swapoff().

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] sparc32 cpumask bitop build fix
Paul Jackson [Mon, 5 Jul 2004 03:25:57 +0000 (20:25 -0700)]
[PATCH] sparc32 cpumask bitop build fix

With the following changes, I was able to compile the "sparc" (32 bit)
arch, using defconfig and crosstool.  There were still plenty of warnings,
but nothing else relating to bitops or cpumasks that I noticed.  This is
working with 2.6.7-mm5.

I have no way to boot test this, but these changes seem obvious enough that
I'd recommend including them.

Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] Fix ia64 UPF_RESOURCES pcdp.c 2.6.7-mm5 build
Paul Jackson [Mon, 5 Jul 2004 03:25:46 +0000 (20:25 -0700)]
[PATCH] Fix ia64 UPF_RESOURCES pcdp.c 2.6.7-mm5 build

It looks like someone removed UPF_RESOURCES in remove-upf_resources.patch
in parallel with someone adding drivers/firmware/pcdp.c that references
UPF_RESOURCES.

In any event, trying to build a defconfig ia64 2.6.7-mm5 (which includes
CONFIG_SERIAL_8250=y in the .config) fails with:

drivers/firmware/pcdp.c: In function `setup_serial_console':
drivers/firmware/pcdp.c:100: error: `UPF_RESOURCES' undeclared (first use in this function)

Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err1-7, err1-8: double locking fix for radeonfb
David Eger [Mon, 5 Jul 2004 03:25:35 +0000 (20:25 -0700)]
[PATCH] err1-7, err1-8: double locking fix for radeonfb

radeonfb: the Stanford lock checker found us double-locking rinfo->reg_lock
via sequences like OUTPLL(foo, INPLL(bar) | baz ), as both OUTPLL and INPLL
grab the register lock.  This should fix the problem.

Signed-off-by: David Eger <eger@havoc.gtf.org
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] selinux space saving
Chris Wright [Mon, 5 Jul 2004 03:25:23 +0000 (20:25 -0700)]
[PATCH] selinux space saving

Move inline avc_cache_stats helper functions alongside the statically declared
avc_cache_stats array.

Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] selinux build fix
Chris Wright [Mon, 5 Jul 2004 03:25:12 +0000 (20:25 -0700)]
[PATCH] selinux build fix

With gcc-3.5 from CVS:

security/selinux/avc.c:76: error: static declaration of 'avc_cache_stats' follows non-static declaration
security/selinux/include/avc.h:107: error: previous declaration of 'avc_cache_stats' was here

Signed-off-by: Chris Wright <chrisw@osdl.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] MTRR __initdata fix
Dave Jones [Mon, 5 Jul 2004 03:25:01 +0000 (20:25 -0700)]
[PATCH] MTRR __initdata fix

smp_changes_mask is used by generic_set_all() which isn't __init

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] wavefront_fx.c build fix
Andrew Morton [Mon, 5 Jul 2004 03:24:50 +0000 (20:24 -0700)]
[PATCH] wavefront_fx.c build fix

With gcc from CVS:

sound/isa/wavefront/wavefront_fx.c:697: error: static declaration of 'page_zero' follows non-static declaration
include/sound/yss225.h:4: error: previous declaration of 'page_zero' was here

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] fix up physnode_map
Martin J. Bligh [Mon, 5 Jul 2004 03:24:39 +0000 (20:24 -0700)]
[PATCH] fix up physnode_map

At the moment, we initialise physnode_map from the various BIOS tables,
which can create problems, as holes inside an node return 1 for pfn_valid,
and yet pfn_to_nid is not correct for them.  I'd hacked around this in my
tree by defaulting the mapping to 0, not -1, but that's not the correct fix
...  this is.

I consolidated all the code back into 1 place, and use node_start_pfn[] and
node_end_pfn[] to walk over it instead - that means it matches up perfectly
with lmem_map's as we're using the same data.  It also cleans up a lot of
the code.

Tested on both NUMA-Q and x440 ...  and it only affects i386 NUMA boxen.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] Fix sparse warnings in fs/udf/*
Mika Kukkonen [Mon, 5 Jul 2004 03:24:28 +0000 (20:24 -0700)]
[PATCH] Fix sparse warnings in fs/udf/*

  CHECK   fs/udf/dir.c
fs/udf/dir.c:240:13: warning: expected lvalue for member dereference
[...]
  CHECK   fs/udf/namei.c
fs/udf/namei.c:872:6: warning: expected lvalue for member dereference
fs/udf/namei.c:916:6: warning: expected lvalue for member dereference
fs/udf/namei.c:1189:14: warning: expected lvalue for member dereference
fs/udf/namei.c:1234:7: warning: expected lvalue for member dereference

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] port reboot workarounds to new DMI probing
Andrey Panin [Mon, 5 Jul 2004 03:24:16 +0000 (20:24 -0700)]
[PATCH] port reboot workarounds to new DMI probing

This patch moves reboot related workarounds out of dmi_scan.c Please
consider applying.

Signed-off-by: Andrey Panin <pazke@donpac.ru>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: remove MachineCheck_Pseries
Andrew Morton [Mon, 5 Jul 2004 03:24:05 +0000 (20:24 -0700)]
[PATCH] ppc64: remove MachineCheck_Pseries

rmk added a check to vmlinux for undefined symbols in -mm kernels.  ppc64
fails due to trying to export an unimplemented symbol from assembly code.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc32: biarch gcc support
Olaf Hering [Mon, 5 Jul 2004 03:23:54 +0000 (20:23 -0700)]
[PATCH] ppc32: biarch gcc support

A native powerpc64-linux gcc can not compile a ppc32 kernel properly.  This
patch fixes it.  It was copied from ppc64.  The change to vmlinux.lds.S
fixes this error:

ld: warning: powerpc:common architecture of input file `init/built-in.o' is
incompatible with powerpc:common64 output

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: SPLPAR spinlock optimisation
Anton Blanchard [Mon, 5 Jul 2004 03:23:43 +0000 (20:23 -0700)]
[PATCH] ppc64: SPLPAR spinlock optimisation

Currently our spinlocks can call into the hypervisor on dedicated processor
machines.  Doing this may slow our locks down, so avoid it where possible.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: gcc 3.5 fixes #2
Anton Blanchard [Mon, 5 Jul 2004 03:23:31 +0000 (20:23 -0700)]
[PATCH] ppc64: gcc 3.5 fixes #2

Here are the ppc64 specific gcc 3.5 fixes.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: gcc 3.5 fixes
Anton Blanchard [Mon, 5 Jul 2004 03:23:20 +0000 (20:23 -0700)]
[PATCH] ppc64: gcc 3.5 fixes

gcc 3.5 ICEd on the ppc64 version of __ptep_set_access_flags, but it does
look like we want to use *ptep here.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: RTAS error log locking fix
Paul Mackerras [Mon, 5 Jul 2004 03:23:09 +0000 (20:23 -0700)]
[PATCH] ppc64: RTAS error log locking fix

From: Linas Vepstas <linas@austin.ibm.com>

When an RTAS call returns the "hardware error" code, we need to do another
RTAS call to find out what went wrong.  Previously we weren't doing that
inside the lock that serializes RTAS calls, and thus another cpu could get
in and do another RTAS call in the meantime.  This patch fixes it.  This
patch also includes some minor whitespace fixes.

Signed-off-by: Linas Vepstas <linas@linas.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: EEH fixes for POWER5 machines (2/2)
Paul Mackerras [Mon, 5 Jul 2004 03:22:58 +0000 (20:22 -0700)]
[PATCH] ppc64: EEH fixes for POWER5 machines (2/2)

From: Linas Vepstas <linas@austin.ibm.com>

This patch fixes the usage of the slot-error-detail log buffer for the
Power5 architecture.  The size of the error buffer is variable, and the
correct size to use should have been obtained from firmware.  Failure to
use the correct buffer sizes will result in hard-to-debug system lockups
deep in firmware.  This patch is based on an earlier patch from Ben
Herrenschmidt, which essentially did the same thing.

This patch also tweaks some of the subroutine documentation.

Signed-off-by: Linas Vepstas <linas@linas.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: EEH fixes for POWER5 machines (1/2)
Paul Mackerras [Mon, 5 Jul 2004 03:22:47 +0000 (20:22 -0700)]
[PATCH] ppc64: EEH fixes for POWER5 machines (1/2)

From: Linas Vepstas <linas@austin.ibm.com>

This patch allows ppc64 to boot on Power5 machines.  The new Power5 PCI
bridge design requires EEH (enhanced PCI error handling) to be enabled for
all PCI devices, not just some PCI devices.  In addition, this patch moves
the check for PCI to ISA bridges out of perf critical code, and into
initialization code.  This also avoids race conditions where the device
type might not have been set.  Also, some whitespace fixes, and some
error-message-printing beautification.

Signed-off-by: Linas Vepstas <linas@linas.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] v850: Return value from no_action in irq.c
Miles Bader [Mon, 5 Jul 2004 03:12:24 +0000 (20:12 -0700)]
[PATCH] v850: Return value from no_action in irq.c

Signed-off-by: Miles Bader <miles@gnu.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] v850: Use __volatile__ qualifier on test_bit asm statements
Miles Bader [Mon, 5 Jul 2004 03:12:13 +0000 (20:12 -0700)]
[PATCH] v850: Use __volatile__ qualifier on test_bit asm statements

Otherwise recent versions of gcc seem to optimize away some necessary tests.

Signed-off-by: Miles Bader <miles@gnu.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] v850: Get rid of lvalue-casts in memset.c to make gcc happy
Miles Bader [Mon, 5 Jul 2004 03:12:02 +0000 (20:12 -0700)]
[PATCH] v850: Get rid of lvalue-casts in memset.c to make gcc happy

Signed-off-by: Miles Bader <miles@gnu.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[Bluetooth] Allocate protocol number for AVDTP support
Marcel Holtmann [Sun, 4 Jul 2004 19:28:19 +0000 (21:28 +0200)]
[Bluetooth] Allocate protocol number for AVDTP support

This patch allocates the next free protocol number for the upcoming
AVDTP support.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
22 years ago[Bluetooth] Add HID protocol support
Marcel Holtmann [Sun, 4 Jul 2004 18:42:38 +0000 (20:42 +0200)]
[Bluetooth] Add HID protocol support

This patch adds support for the Bluetooth HID protocol to the
Bluetooth subsystem. Currently only the boot mode is supported.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
22 years ago[Bluetooth] Fix kobject oops on firmware loading
Marcel Holtmann [Sun, 4 Jul 2004 16:55:47 +0000 (18:55 +0200)]
[Bluetooth] Fix kobject oops on firmware loading

Allocate the memory for .kobj.k_name and make use of kobject_init() for
the fake device that is needed for firmware loading.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
22 years ago[Bluetooth] Fix deadlock in the 3Com driver
Marcel Holtmann [Sun, 4 Jul 2004 16:46:07 +0000 (18:46 +0200)]
[Bluetooth] Fix deadlock in the 3Com driver

This patch fixes the deadlock problem when bt3c_interrupt() is calling
bt3c_write_wakeup() with info->lock held.

Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
22 years ago[Bluetooth] Change inquiry_cache to hci_inquiry_cache
Marcel Holtmann [Sun, 4 Jul 2004 14:58:09 +0000 (16:58 +0200)]
[Bluetooth] Change inquiry_cache to hci_inquiry_cache

This patch changes the inquiry_* functions to be either local to hci_core.c
or change name from inquiry_cache_lookup to hci_inquiry_cache_lookup to try
and reduce the potential for namespace conflicts. There could be USB or SCSI
or other code that uses an inquiry_cache.

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: Marcel Holtmann <marcel@holtmann.org>
22 years ago[PATCH] swap_unplug_io_fn() nommu update
Paul Mundt [Sun, 4 Jul 2004 05:01:12 +0000 (22:01 -0700)]
[PATCH] swap_unplug_io_fn() nommu update

include/linux/swap.h changed the definition for swap_unplug_io_fn() awhile
back, but mm/nommu.c was never updated to reflect the new definition.
As such, mm/nommu.c presently fails to compile. This fixes it.

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc32: compilation failure on ppc32
Christoph Hellwig [Sun, 4 Jul 2004 05:00:37 +0000 (22:00 -0700)]
[PATCH] ppc32: compilation failure on ppc32

This fixes compilation on ppc32.

The power/smp.o file should be linked only if both SMP and SWSUSPEND
are configured in. It used to do it even without SWSUSPEND.

22 years agoMerge bk://gkernel.bkbits.net/libata-2.6
Linus Torvalds [Sun, 4 Jul 2004 04:57:22 +0000 (21:57 -0700)]
Merge bk://gkernel.bkbits.net/libata-2.6
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years ago[libata sata_promise] update driver to use new ->qc_issue hook
Jeff Garzik [Sat, 3 Jul 2004 22:12:44 +0000 (18:12 -0400)]
[libata sata_promise] update driver to use new ->qc_issue hook

The ->qc_issue hook was designed to allow drivers to override some
or all of the actual delivery of the taskfile to hardware.

In the case of Promise, the hardware has its own packet format when doing
read/write DMA commands, but uses traditional ATA taskfile registers
for other types of commands.

22 years ago[libata] create, and use aga_sg_init[_one] helpers
Jeff Garzik [Sat, 3 Jul 2004 20:57:38 +0000 (16:57 -0400)]
[libata] create, and use aga_sg_init[_one] helpers

Move more SCSI-specific code out of libata core.

22 years ago[libata] split ATA_QCFLAG_SG into ATA_QCFLAG_{SG,SINGLE}
Jeff Garzik [Sat, 3 Jul 2004 20:26:17 +0000 (16:26 -0400)]
[libata] split ATA_QCFLAG_SG into ATA_QCFLAG_{SG,SINGLE}

In part of the effort to remove SCSI specifics from the libata
internals, remove references to cmd->use_sg.  cmd->use_sg
becomes ATA_QCFLAG_SG, and !cmd->use_sg becomes ATA_QCFLAG_SINGLE.

Convenience constant ATA_QCFLAG_DMAMAP is created when the programmer
wishes to refer collectively to ATA_QCFLAG_{SG,SINGLE}.

22 years ago[ata] add ata_ok() inlined helper, and ATA_{DRDY,DF} bit to linux/ata.h
Jeff Garzik [Sat, 3 Jul 2004 18:42:40 +0000 (14:42 -0400)]
[ata] add ata_ok() inlined helper, and ATA_{DRDY,DF} bit to linux/ata.h

22 years ago[libata] create, and use, ->irq_clear hook
Jeff Garzik [Sat, 3 Jul 2004 18:40:48 +0000 (14:40 -0400)]
[libata] create, and use, ->irq_clear hook

This is more conservative in general, and so applies to multiple
controllers.  Specifically it attempts to address irq-related issues
on the Intel ICH5/6 hardware.  On Intel ICH5/6, the BMDMA 'interrupt'
status bit will be set even on non-DMA commands, which software
(and I) did not expect.

This change clears pending interrupts once upon initialization,
and then each time ata_irq_on() is called.

22 years ago[PATCH] remaining cpumask const qualifiers
Paul Jackson [Sat, 3 Jul 2004 16:55:19 +0000 (09:55 -0700)]
[PATCH] remaining cpumask const qualifiers

The remainder of the const qualifiers on cpumask ops.

My cpumask overhaul missed specifying the const qualifiers
in cpumask.h.  Subsequently, Linus has added some.  The following
should provide the remainder of them.  It also fixes one src vs dst
variable misnaming.

Using crosstool on 2.6.7-mm5, I have built the following
arch's with the following change included:

  alpha ia64 powerpc-405 powerpc-750 sparc sparc64 x86_64

Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years agoMerge bk://bk.arm.linux.org.uk/linux-2.6-serial
Linus Torvalds [Sat, 3 Jul 2004 16:51:56 +0000 (09:51 -0700)]
Merge bk://bk.arm.linux.org.uk/linux-2.6-serial
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years agoMerge bk://bk.arm.linux.org.uk/linux-2.6-rmk
Linus Torvalds [Sat, 3 Jul 2004 16:48:45 +0000 (09:48 -0700)]
Merge bk://bk.arm.linux.org.uk/linux-2.6-rmk
into ppc970.osdl.org:/home/torvalds/v2.6/linux

22 years ago[PATCH] apm.c warning fix
Andrew Morton [Sat, 3 Jul 2004 03:05:24 +0000 (20:05 -0700)]
[PATCH] apm.c warning fix

arch/i386/kernel/apm.c: In function `suspend':
arch/i386/kernel/apm.c:1221: warning: implicit declaration of function `save_processor_state'
arch/i386/kernel/apm.c:1223: warning: implicit declaration of function `restore_processor_state'

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] s/2.5/2.6/ in MAINTAINERS
Adrian Bunk [Sat, 3 Jul 2004 03:05:12 +0000 (20:05 -0700)]
[PATCH] s/2.5/2.6/ in MAINTAINERS

s/2.5/2.6/ in MAINTAINERS

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] telephony: support devfs
Paul King [Sat, 3 Jul 2004 03:04:58 +0000 (20:04 -0700)]
[PATCH] telephony: support devfs

Add devfs support to telephony devices.  Only tested with a single
telephony device.  Devices name phone/<minor>.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] dma_get_required_mask() build fix
James Bottomley [Sat, 3 Jul 2004 03:04:47 +0000 (20:04 -0700)]
[PATCH] dma_get_required_mask() build fix

This new function fails to build on sparc64 due to nasty include
dependencies.

Fix that by uninlining it - it was too big for inlining anyway.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err2-29: ufs_new_fragments() locking fix
Andrew Morton [Sat, 3 Jul 2004 03:04:35 +0000 (20:04 -0700)]
[PATCH] err2-29: ufs_new_fragments() locking fix

Found by the Stanford locking checker

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err2-27: i2o_claim_device() locking fix
Andrew Morton [Sat, 3 Jul 2004 03:04:24 +0000 (20:04 -0700)]
[PATCH] err2-27: i2o_claim_device() locking fix

Found by the Stanford locking checker

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err2-25: dvb_register_i2c_bus() locking fix
Andrew Morton [Sat, 3 Jul 2004 03:04:12 +0000 (20:04 -0700)]
[PATCH] err2-25: dvb_register_i2c_bus() locking fix

Found by the Stanford locking checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err2-1 dvb_register_i2c_device locking fix
Andrew Morton [Sat, 3 Jul 2004 03:03:59 +0000 (20:03 -0700)]
[PATCH] err2-1 dvb_register_i2c_device locking fix

Fix deadlock identified by the Stanford locking checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err1-14: sb_audio locking fix
Andrew Morton [Sat, 3 Jul 2004 03:03:48 +0000 (20:03 -0700)]
[PATCH] err1-14: sb_audio locking fix

Lamely fix a straightforward deadlock in sb_audio.c.  Founf by the Stanford
locking checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err1-2: sscape locking fix
Andrew Morton [Sat, 3 Jul 2004 03:03:36 +0000 (20:03 -0700)]
[PATCH] err1-2: sscape locking fix

Fix deadlock identified by the Stanford locking checker.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] err1-40: sysvfs locking fix
Andrew Morton [Sat, 3 Jul 2004 03:03:25 +0000 (20:03 -0700)]
[PATCH] err1-40: sysvfs locking fix

Found by the new Stanford locking checker.

Minimal fix for a deadlock in sysvfs: get_branch() can take
read_lock(&pointers_lock), but one caller already has a write_lock.

Perhaps some of the "oh we raced, drop everything and try again" logic in
there can go away now, but this is enugh to fix the obvious deadlock.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] fbcon: optimization for accel_putcs()
Antonino Daplas [Sat, 3 Jul 2004 03:03:13 +0000 (20:03 -0700)]
[PATCH] fbcon: optimization for accel_putcs()

I did some simple benchmarking (time cat linux-2.6.7-mm5/MAINTAINERS)
between 2.4 and 2.6 and I am not satisfied with what I see (It's claimed
that fbdev-2.6 is faster than 2.4).  The reason for the claim:

2.4 putcs - draw small amounts of data a lot of times
2.6 putcs - draw larger amounts of data a fewer times

The way characters are drawn in 2.6 is optimal for accelerated drivers but
should also give a speed boost for drivers that rely on software drawing.
However the penaly incurred when preparing a large bitmap from a number of
small bitmaps is currently very high.  This is because of the following
reasons:

1 fb_move_buf_{aligned|unaligned} uses pixmap->{out|in}buf.  This is very
  expensive since outbuf and inbuf methods process only a byte or 2 of data
  at a time.

2 fb_sys_outbuf (the default method for pixmap->outbuf) uses memcpy().
  Not a good choice if moving only a few bytes.

3 fb_move_buf_unaligned (used for fonts such as 12x22) also involves a
  lot of bit operations + a lot of calls to outbuf/inbuf which
  proportionately increases the penaly.

So, I thought of separating fb_move_buf_* to fb_iomove_buf_* and
fb_sysmove_buf_*.

fb_iomove_buf_* - used if drivers specified outbuf and inbuf methods
fb_sysmove_buf_* - used if drivers have no outbuf or inbuf methods

*Most, if not all drivers fall in the second category.

Below is a table that show differences between 2.4, 2.6 and 2.6 +
abovementioned changes.  To reduce the effect of panning and
fillrect/copyarea, the scrollmode is forced to redraw.

=================================================================
Test Hardware: P4 2G nVidia GeForce2 MX 64
Scrollmode: redraw

time cat linux-2.6.7-mm5/MAINTAINERS

1024x768-8 1024x768-16 1024x768-32
=================================================================
8x16 noaccel (2.4)
real    0m5.490s real    0m8.535s real    0m15.388s
user    0m0.001s user    0m0.000s user    0m0.001s
sys     0m5.487s sys     0m8.535s sys     0m15.386s

8x16 noaccel (2.6)
real    0m5.166s real    0m7.195s real    0m12.177s
user    0m0.001s user    0m0.000s user    0m0.000s
sys     0m5.164s sys     0m7.192s sys     0m12.176s

8x16 noaccel+patch (2.6)
real    0m3.474s real    0m5.496s real    0m10.460s
user    0m0.001s user    0m0.001s user    0m0.001s
sys     0m5.492s sys     0m5.492s sys     0m10.454s
=================================================================
8x16 accel (2.4)
real    0m4.368s real    0m9.420s real    0m22.415s
user    0m0.001s user    0m0.001s user    0m0.001s
sys     0m4.019s sys     0m9.384s sys     0m22.312s

8x16 accel (2.6)
real    0m4.296s real    0m4.339s real    0m4.391s
user    0m0.001s user    0m0.001s user    0m0.000s
sys     0m4.280s sys     0m4.336s sys     0m4.389s

8x16 accel+patch (2.6)
real    0m2.536s real    0m2.649s real    0m2.799s
user    0m0.000s user    0m0.000s user    0m0.001s
sys     0m2.536s sys     0m2.645s sys     0m2.798s
=================================================================

1024x768-8 1024x768-16 1024x768-32
=================================================================
12x22 noaccel (2.4)
real    0m7.883s real    0m12.175s real    0m21.134s
user    0m0.000s user    0m0.000s user    0m0.001s
sys     0m7.882s sys     0m12.174s sys     0m21.129s

12x22 noaccel (2.6)
real    0m10.651s real    0m13.550s real    0m21.009s
user    0m0.001s user    0m0.001s user    0m0.000s
sys     0m10.617s sys     0m13.545s sys     0m21.008s

12x22 noaccel+patch (2.6)
real    0m4.794s real    0m7.718s real    0m15.173s
user    0m0.002s user    0m0.001s user    0m0.000s
sys     0m4.792s sys     0m7.715s sys     0m15.170s
=================================================================
12x22 accel (2.4)
real    0m3.971s real    0m9.030s real    0m21.711s
user    0m0.000s user    0m0.000s user    0m0.000s
sys     0m3.950s sys     0m8.983s sys     0m21.602s

12x22 accel (2.6)
real    0m9.392s real    0m9.486s real    0m9.508s
user    0m0.000s user    0m0.000s user    0m0.001s
sys     0m9.392s sys     0m9.484s sys     0m9.484s

12x22 accel+patch (2.6)
real    0m3.570s real    0m3.603s real    0m3.848s
user    0m0.001s user    0m0.000s user    0m0.000s
sys     0m3.567s sys     0m3.600s sys     0m3.844s
=================================================================

Summary:

1 2.6 unaccelerated is a bit faster than 2.4 when handling 8x16 fonts,
  with a higher speed differential at high color depths.

2 2.4 unaccelerated is a bit faster than 2.6 when handling 12x22 fonts,
  with a smaller speed difference at high color depths (2.6 is actually a
  bit faster than 2.4 at 32bpp).

3 2.4 rivafb accelerated suffers at high color depths, even becoming
  slower than unaccelerated, possibly because of the 'draw few bytes many
  times' method.

4 2.6 rivafb accelerated has similar performance at any color depth,
  possibly because of 'draw lots of bytes a fewer times' method.

5 With the changes, there is a speed gain of ~1.7 seconds and ~5.7
  seconds with 8x16 and 12x22 fonts respectively indepependent of the color
  depth or acceleration used.  The speed gain is constant but significant.

Below is a patch against 2.6.7-mm5.  The effects will be very noticeable
with drivers that uses SCROLL_REDRAW, but one should still see some speed
gain even if SCROLL_YPAN/YWRAP is used.

Separated fb_sys_move_* into fb_iosys_move_* and fb_sysmove_* to reduce
penalty when constructing fb_image->data from character maps.  In my
testcase (1024x768 SCROLL_REDRAW), I get a ~1.7 second advantage with 'time
cat MAINTAINERS' using 8x16 fonts and ~5.7 seconds with 12x22 fonts.  The
speed gain is independent of acceleration or color depth.

Signed-off-by: Antonino Daplas <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] remove unused variable in mxser.c
Adrian Bunk [Sat, 3 Jul 2004 03:03:02 +0000 (20:03 -0700)]
[PATCH] remove unused variable in mxser.c

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] remove unused variable in esp.c
Adrian Bunk [Sat, 3 Jul 2004 03:02:50 +0000 (20:02 -0700)]
[PATCH] remove unused variable in esp.c

Signed-off-by: Adrian Bunk <bunk@fs.tum.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] i386 math emu build fix
Andrew Morton [Sat, 3 Jul 2004 03:02:39 +0000 (20:02 -0700)]
[PATCH] i386 math emu build fix

rmk's patch found a defined-but-unimplemented symbol

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] Fix minor quota race
Jan Kara [Sat, 3 Jul 2004 03:02:27 +0000 (20:02 -0700)]
[PATCH] Fix minor quota race

It fixes a possible race between quotaoff and prune_icache.  The race could
lead to some forgotten pointers to quotas in inodes leading later to BUG
when invalidating quota structures.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ia32 NUMA: physnode_map entries can be negative
Martin J. Bligh [Sat, 3 Jul 2004 03:02:16 +0000 (20:02 -0700)]
[PATCH] ia32 NUMA: physnode_map entries can be negative

Based on work from Bill Irwin <wli@holomorphy.com>

physnode_map[] needs to be signed so that pfn_to_nid() can return negative
values used to detect invalid pfn's.

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc32: Redwood[56] support for smc91x Ethernet driver
Matt Porter [Sat, 3 Jul 2004 03:02:05 +0000 (20:02 -0700)]
[PATCH] ppc32: Redwood[56] support for smc91x Ethernet driver

This patch enables the Redwood 5 and Redwood 6 platforms to use the smc91x
ethernet driver.

Signed-off-by: Dale Farnsworth <dale@farnsworth.org>
Signed-off-by: Matt Porter <mporter@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: set ppc_md.log_error
Paul Mackerras [Sat, 3 Jul 2004 03:01:53 +0000 (20:01 -0700)]
[PATCH] ppc64: set ppc_md.log_error

We have a platform-specific function pointer on ppc64 for a function to log
errors detected by the platform, but it was never getting set.  This patch
sets it on pSeries (the only ppc64 platform which has an error logging
function).

Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ppc64: remove deprecated firmware API
Paul Mackerras [Sat, 3 Jul 2004 03:01:42 +0000 (20:01 -0700)]
[PATCH] ppc64: remove deprecated firmware API

From: Linas Vepstas <linas@austin.ibm.com>

This patch eliminates the usage of the deprecated ibm,fw-phb-id token for
idnetifying PCI bus heads in favor of the documented, offically supported
mechanism for obtaining this info.  Please note that some versions of
firmware may return incorrect values for the ibm,fw-phb-id token.

Signed-off-by: Linas Vepstas <linas@linas.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[PATCH] ntfs build fix
Andrew Morton [Sat, 3 Jul 2004 03:01:30 +0000 (20:01 -0700)]
[PATCH] ntfs build fix

gcc-2.95 chokes on this.

Cc: Anton Altaparmakov <aia21@cantab.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
22 years ago[libata sata_sil] add drive to mod15write quirk list
Ricky Beam [Fri, 2 Jul 2004 22:19:47 +0000 (18:19 -0400)]
[libata sata_sil] add drive to mod15write quirk list

22 years ago[libata] add ata_queued_cmd completion hook
Jeff Garzik [Fri, 2 Jul 2004 20:23:24 +0000 (16:23 -0400)]
[libata] add ata_queued_cmd completion hook

22 years ago[libata] add ->qc_issue hook
Jeff Garzik [Fri, 2 Jul 2004 20:08:29 +0000 (16:08 -0400)]
[libata] add ->qc_issue hook

This hook is used when an ATA controller wishes to use
hardware-specific methods of taskfile delivery, rather
than the standard method of bitbanging the ATA shadow
registers.

22 years ago[PATCH] chown permission check fix for ATTR_GID
Chris Wright [Fri, 2 Jul 2004 13:55:04 +0000 (06:55 -0700)]
[PATCH] chown permission check fix for ATTR_GID

SuSE discovered this problem with chown and ATTR_GID.  Make sure user
is authorized to change the group, CAN-2004-0497.