deliverable/linux.git
18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Thu, 26 Oct 2006 14:44:07 +0000 (07:44 -0700)] 
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [BRIDGE]: correct print message typo
  [TCP] H-TCP: fix integer overflow
  [TCP] cubic: scaling error

18 years ago[BRIDGE]: correct print message typo
Randy Dunlap [Thu, 26 Oct 2006 06:07:37 +0000 (23:07 -0700)] 
[BRIDGE]: correct print message typo

Correct message typo/spello.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[TCP] H-TCP: fix integer overflow
Gavin McCullagh [Thu, 26 Oct 2006 06:05:52 +0000 (23:05 -0700)] 
[TCP] H-TCP: fix integer overflow

When using H-TCP with a single flow on a 500Mbit connection (or less
actually), alpha can exceed 65000, so alpha needs to be a u32.

Signed-off-by: Gavin McCullagh <gavin.mccullagh@nuim.ie>
Signed-off-by: Doug Leith <doug.leith@nuim.ie>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[TCP] cubic: scaling error
Stephen Hemminger [Thu, 26 Oct 2006 06:04:12 +0000 (23:04 -0700)] 
[TCP] cubic: scaling error

Doug Leith observed a discrepancy between the version of CUBIC described
in the papers and the version in 2.6.18. A math error related to scaling
causes Cubic to grow too slowly.

Patch is from "Sangtae Ha" <sha2@ncsu.edu>. I validated that
it does fix the problems.

See the following to show behavior over 500ms 100 Mbit link.

Sender (2.6.19-rc3) ---  Bridge (2.6.18-rt7) ------- Receiver (2.6.19-rc3)
                    1G      [netem]           100M

http://developer.osdl.org/shemminger/tcp/2.6.19-rc3/cubic-orig.png
http://developer.osdl.org/shemminger/tcp/2.6.19-rc3/cubic-fix.png

Signed-off-by: Stephen Hemminger <shemminger@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years agoMerge branch 'upstream-fixes' of git://lost.foo-projects.org/~ahkok/git/netdev-2...
Jeff Garzik [Thu, 26 Oct 2006 04:11:28 +0000 (21:11 -0700)] 
Merge branch 'upstream-fixes' of git://lost.foo-projects.org/~ahkok/git/netdev-2.6 into upstream-fixes

18 years ago[PATCH] AVR32: Update defconfig
Haavard Skinnemoen [Tue, 24 Oct 2006 08:12:46 +0000 (10:12 +0200)] 
[PATCH] AVR32: Update defconfig

Sync atstk1002_defconfig with latest git, turn off non-existent
drivers and enable a few more userspace-visible options like
SysV IPC and inotify support.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AVR32: Use __raw MMIO access for internal peripherals
Haavard Skinnemoen [Tue, 24 Oct 2006 08:12:45 +0000 (10:12 +0200)] 
[PATCH] AVR32: Use __raw MMIO access for internal peripherals

The read[bwl] and write[bwl] functions are meant for accessing PCI
devices. How this is achieved on AVR32 is unknown, as there are no
systems with a PCI bridge available yet.

On-chip peripheral access, however, should not depend on how we end
up implementing PCI access, so using __raw_read[bwl]/__raw_write[bwl]
is the right thing to do for on-chip peripherals. This patch converts
the drivers for the static memory controller, interrupt controller,
PIO controller and system manager to use __raw MMIO access.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AVR32: Implement and export __raw_{read,write}s[bwl]
Haavard Skinnemoen [Tue, 24 Oct 2006 08:12:44 +0000 (10:12 +0200)] 
[PATCH] AVR32: Implement and export __raw_{read,write}s[bwl]

Implement __raw_readsb and __raw_writesb. Export __raw_reads[bwl]
and __raw_writes[bwl] for use by modules.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AVR32: add io{read,write}{8,16,32}{be,} support
Ben Nizette [Tue, 24 Oct 2006 08:12:43 +0000 (10:12 +0200)] 
[PATCH] AVR32: add io{read,write}{8,16,32}{be,} support

A number of new drivers require io{read,write}{8,16,32}{be,} family of io
operations.  These are provided for the AVR32 by this patch in the form of
a series of macros.

Access to the (memory mapped) io space through these macros is defined to
be little endian only as little endian devices (such as PCI) are the main
consumer of IO access.  If high speed access is required,
io{read,write}{16,32}be macros are supplied to perform native big endian
access to this io space.

Signed-off-by: Ben Nizette <ben@mallochdigital.com>
Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AVR32: Fix oversize immediates in atomic.h
Haavard Skinnemoen [Tue, 24 Oct 2006 08:12:42 +0000 (10:12 +0200)] 
[PATCH] AVR32: Fix oversize immediates in atomic.h

When calling e.g. atomic_sub_return with a large constant, the
compiler may output an immediate that is too large for the sub
instruction in the middle of the loop.

Fix this by explicitly specifying the number of bits allowed in the
constraint. Also stop atomic_add_return() and friends from falling
back to their respective "sub" variants if the constant is too large
to fit in an immediate.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AVR32: Don't try to iounmap P2 segment addresses
Haavard Skinnemoen [Tue, 24 Oct 2006 08:12:41 +0000 (10:12 +0200)] 
[PATCH] AVR32: Don't try to iounmap P2 segment addresses

While ioremap() will happily map a physical address through the
P2 (uncached) segment when appropriate, iounmap() doesn't know how
to handle those mappings.

This patch makes iounmap() do the right thing, i.e. nothing, for
such mappings.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AVR32: Silence some compile warnings
Haavard Skinnemoen [Tue, 24 Oct 2006 08:12:40 +0000 (10:12 +0200)] 
[PATCH] AVR32: Silence some compile warnings

Silence a few compile warnings which are basically harmless, but
easy to fix.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] AVR32: Minor Makefile cleanup
Haavard Skinnemoen [Tue, 24 Oct 2006 08:12:39 +0000 (10:12 +0200)] 
[PATCH] AVR32: Minor Makefile cleanup

Don't generate listing by default, remove unused LIBGCC variable and
rename generated disassembly and listing files to vmlinux.{s,lst}.

Also make sure that files generated during the build are actually
removed with make clean.

Signed-off-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6
Linus Torvalds [Thu, 26 Oct 2006 03:22:55 +0000 (20:22 -0700)] 
Merge /linux/kernel/git/jejb/scsi-rc-fixes-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6: (40 commits)
  [SCSI] aic79xx: Print out signalling
  [SCSI] aic7xxx: Remove slave_destroy
  [SCSI] aic79xx: set precompensation
  [SCSI] aic79xx: Fixup external device reset
  [SCSI] replace u8 and u32 with __u8 and __u32 in scsi.h for user space
  [SCSI] lpfc: fix printk format warning
  [SCSI] aic79xx: make ahd_set_tags() static
  [SCSI] aic7xxx: cleanups
  [SCSI] drivers/scsi: Handcrafted MIN/MAX macro removal
  [SCSI] scsi_debug: support REPORT TARGET PORT GROUPS
  [SCSI] qla1280 bus reset typo
  [SCSI] libiscsi: fix logout pdu processing
  [SCSI] libiscsi: fix aen support
  [SCSI] libiscsi: fix missed iscsi_task_put in xmit error path
  [SCSI] libiscsi: fix oops in connection create failure path
  [SCSI] iscsi class: fix slab corruption during restart
  [SCSI] Switch fdomain to the pci_get API
  [SCSI] add can_queue to host parameters
  [SCSI] megaraid_{mm,mbox}: 64-bit DMA capability fix
  [SCSI] aic94xx: Supermicro motherboards support
  ...

18 years agoMerge branch 'intelfb-patches' of master.kernel.org:/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Thu, 26 Oct 2006 03:22:04 +0000 (20:22 -0700)] 
Merge branch 'intelfb-patches' of /linux/kernel/git/airlied/intelfb-2.6

* 'intelfb-patches' of master.kernel.org:/pub/scm/linux/kernel/git/airlied/intelfb-2.6:
  Remove unnecessary check in drivers/video/intelfb/intelfbhw.c
  intel fb: switch to pci_get API

18 years agoMerge branch 'drm-patches' of master.kernel.org:/pub/scm/linux/kernel/git/airlied...
Linus Torvalds [Thu, 26 Oct 2006 03:21:43 +0000 (20:21 -0700)] 
Merge branch 'drm-patches' of /linux/kernel/git/airlied/drm-2.6

* 'drm-patches' of master.kernel.org:/pub/scm/linux/kernel/git/airlied/drm-2.6:
  drm: ioremap balanced with iounmap for drivers/char/drm
  drm: fix error returns, sysfs error handling
  fix return code in error case.
  drm: mga: set dev_priv_size
  drm: savage: dev->agp_buffer_map is not initialized for AGP DMA on savages
  drm: radeon: only allow specific type-3 packetss through verifier

18 years agoMerge branch 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6
Linus Torvalds [Thu, 26 Oct 2006 03:21:16 +0000 (20:21 -0700)] 
Merge branch 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6

* 'for-linus' of git://one.firstfloor.org/home/andi/git/linux-2.6:
  [PATCH] x86-64: Only look at per_cpu data for online cpus.
  [PATCH] x86-64: Simplify the vector allocator.

18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Thu, 26 Oct 2006 03:20:41 +0000 (20:20 -0700)] 
Merge /pub/scm/linux/kernel/git/herbert/crypto-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  [CRYPTO] users: Select ECB/CBC where needed

18 years agoMerge branch 'merge' of master.kernel.org:/pub/scm/linux/kernel/git/paulus/powerpc
Linus Torvalds [Thu, 26 Oct 2006 03:20:03 +0000 (20:20 -0700)] 
Merge branch 'merge' of /linux/kernel/git/paulus/powerpc

* 'merge' of master.kernel.org:/pub/scm/linux/kernel/git/paulus/powerpc:
  [POWERPC] Make sure __cpu_preinit_ppc970 gets called on 970GX processors
  [POWERPC] Fix CHRP platforms with only 8259
  [POWERPC] IPIC: Fix spinlock recursion in set_irq_handler
  [POWERPC] Fix the UCC rx/tx clock of QE
  [POWERPC] cell: update defconfig
  [POWERPC] spufs: fix another off-by-one bug in spufs_mbox_read
  [POWERPC] spufs: fix signal2 file to report signal2
  [POWERPC] Fix device_is_compatible() const warning
  [POWERPC] Cell timebase bug workaround
  [POWERPC] Support feature fixups in modules
  [POWERPC] Support feature fixups in vdso's
  [POWERPC] Support nested cpu feature sections
  [POWERPC] Consolidate feature fixup code
  [POWERPC] Fix hang in start_ldr if _end or _edata is unaligned
  [POWERPC] Fix spelling errors in ucc_fast.c and ucc_slow.c
  [POWERPC] Don't require execute perms on wrapper when building zImage.initrd
  [POWERPC] Add 970GX cputable entry
  [POWERPC] Fix build breakage with CONFIG_PPC32
  [POWERPC] Fix compiler warning message on get_property call
  [POWERPC] Simplify stolen time calculation

18 years agoMerge master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Thu, 26 Oct 2006 03:19:23 +0000 (20:19 -0700)] 
Merge master.kernel.org:/home/rmk/linux-2.6-arm

* master.kernel.org:/home/rmk/linux-2.6-arm:
  [ARM] 3902/1: Enable GPIO81-84 on PXA255
  [ARM] Comment out missing configuration symbols
  [ARM] 3898/1: corgi_bl fix module loading
  [ARM] 3897/1: corgi_bl fix module compiling
  [ARM] Fix breakage in 7281c248f797723f66244b7ecef204620f664648

18 years ago[POWERPC] Make sure __cpu_preinit_ppc970 gets called on 970GX processors
Olof Johansson [Wed, 25 Oct 2006 22:32:40 +0000 (17:32 -0500)] 
[POWERPC] Make sure __cpu_preinit_ppc970 gets called on 970GX processors

Add check for 970GX for __cpu_preinit_ppc970.

Signed-off-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[SCSI] aic79xx: Print out signalling
Hannes Reinecke [Mon, 23 Oct 2006 13:26:37 +0000 (15:26 +0200)] 
[SCSI] aic79xx: Print out signalling

This is a cross-port of a similar patch for aic7xxx;
only it's a bit simpler here as we don't support HVD
and all controller actually implement this register.
I hope.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] aic7xxx: Remove slave_destroy
Hannes Reinecke [Mon, 23 Oct 2006 13:25:36 +0000 (15:25 +0200)] 
[SCSI] aic7xxx: Remove slave_destroy

This is a cross-port from aic79xx; we still hit the occasional
BUG_ON in slave_destroy. And again we don't really need the
slave_destroy callback nor the ahc_linux_target structure
at all.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] aic79xx: set precompensation
Hannes Reinecke [Mon, 23 Oct 2006 13:24:23 +0000 (15:24 +0200)] 
[SCSI] aic79xx: set precompensation

aic79xx has a special 'iocell' chip which handles the precompensation.
If it's set via DV we should make sure to set the chip correctly, too.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] aic79xx: Fixup external device reset
Hannes Reinecke [Mon, 23 Oct 2006 13:22:37 +0000 (15:22 +0200)] 
[SCSI] aic79xx: Fixup external device reset

Whenever an external device is resetted we really have to take
care to keep the channel in sync. Just notifying SCSI-ML and retry
is not enough as we have to make sure the SCSI bus is not getting
confused, either.
So whenever we detect an external reset we rewrite the command to
TUR, disable packetized command and notify the internal engine
that an abort has happened. This way we trigger a proper bus
reset sequence and all devices will be renegotiated properly.
Kudos to Justin Gibbs and Luben Tuikov for this idea.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] replace u8 and u32 with __u8 and __u32 in scsi.h for user space
FUJITA Tomonori [Fri, 20 Oct 2006 00:08:18 +0000 (09:08 +0900)] 
[SCSI] replace u8 and u32 with __u8 and __u32 in scsi.h for user space

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] lpfc: fix printk format warning
Randy Dunlap [Tue, 24 Oct 2006 04:47:13 +0000 (21:47 -0700)] 
[SCSI] lpfc: fix printk format warning

Fix printk format warning:
drivers/scsi/lpfc/lpfc_attr.c:597: warning: long long unsigned int format, uint64_t arg (arg 4)

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: James Smart <James.Smart@Emulex.Com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] aic79xx: make ahd_set_tags() static
Adrian Bunk [Fri, 20 Oct 2006 21:48:07 +0000 (14:48 -0700)] 
[SCSI] aic79xx: make ahd_set_tags() static

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Acked-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] aic7xxx: cleanups
Adrian Bunk [Fri, 20 Oct 2006 21:47:57 +0000 (14:47 -0700)] 
[SCSI] aic7xxx: cleanups

- make needlessly global code static

- #if 0 the following unused global functions:
  - aic79xx_core.c: ahd_print_scb
  - aic79xx_core.c: ahd_suspend
  - aic79xx_core.c: ahd_resume
  - aic79xx_core.c: ahd_dump_scbs
  - aic79xx_osm.c: ahd_softc_comp

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Acked-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] drivers/scsi: Handcrafted MIN/MAX macro removal
Amol Lad [Fri, 20 Oct 2006 21:48:40 +0000 (14:48 -0700)] 
[SCSI] drivers/scsi: Handcrafted MIN/MAX macro removal

Cleanups done to use min/max macros from kernel.h.  Handcrafted MIN/MAX
macros are changed to use macros in kernel.h

[akpm@osdl.org: fix warning]
Signed-off-by: Amol Lad <amol@verismonetworks.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] scsi_debug: support REPORT TARGET PORT GROUPS
Hannes Reinecke [Fri, 20 Oct 2006 07:58:47 +0000 (09:58 +0200)] 
[SCSI] scsi_debug: support REPORT TARGET PORT GROUPS

This patch adds support for REPORT TARGET PORT GROUPS. This is used
eg for the multipathing priority callout to determine the path
priority.
With this patch multipath-tools can use the existing mpath_prio_alua
callout to exercise the path priority grouping.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Douglas Gilbert <dougg@torque.net>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] qla1280 bus reset typo
Jes Sorensen [Tue, 17 Oct 2006 13:32:06 +0000 (09:32 -0400)] 
[SCSI] qla1280 bus reset typo

Fix typo in check of return value of qla1280_bus_reset() which would
result in an adapter reset in addition to the bus reset.

Signed-off-by: Jes Sorensen <jes@sgi.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] libiscsi: fix logout pdu processing
Mike Christie [Mon, 16 Oct 2006 22:09:42 +0000 (18:09 -0400)] 
[SCSI] libiscsi: fix logout pdu processing

According to the iscsi RFC, we cannot send other requests if
we have sent a logout pdu. This patch enforces this requirement
by blocking the session and suspending the send thread. Userspace
decides if we restart the connection or if we just free everything.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] libiscsi: fix aen support
Mike Christie [Mon, 16 Oct 2006 22:09:41 +0000 (18:09 -0400)] 
[SCSI] libiscsi: fix aen support

We have been dropping the pdu. We should just send it to userspace
and let it handle it.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] libiscsi: fix missed iscsi_task_put in xmit error path
Mike Christie [Mon, 16 Oct 2006 22:09:40 +0000 (18:09 -0400)] 
[SCSI] libiscsi: fix missed iscsi_task_put in xmit error path

from bhalevy@gmail.com:

It looks like change 652 to libiscsi.c added some dead code around line
670
                if (rc) {
                        spin_unlock_bh(&conn->session->lock);
                        goto again;
                }

since 5 lines above we goto again if (rc).

It looks like the previous if (rc) should go away if we want to put the
ctask before
breaking out of the while loop with "goto again" (see following patch).

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] libiscsi: fix oops in connection create failure path
Mike Christie [Mon, 16 Oct 2006 22:09:39 +0000 (18:09 -0400)] 
[SCSI] libiscsi: fix oops in connection create failure path

If connection creation fails we end up calling list_del
on a invalid struct. This then causes an oops. We are not
acutally using the lists (old MCS code we thought might
be useful elsewhere) so this patch just removes that
code.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] iscsi class: fix slab corruption during restart
Mike Christie [Mon, 16 Oct 2006 22:09:38 +0000 (18:09 -0400)] 
[SCSI] iscsi class: fix slab corruption during restart

The transport class recv mempools are causing slab corruption.
We could hack around netlink's lack of mempool support like dm,
but it is just too ulgy (dm's hack is ugly enough :) when you need
to support broadcast.

This patch removes the recv pools. We have not used them even when
we were allocting 20 MB per session and the system only had 64 MBs.
And we have no pools on the send side and have been ok there. When
Peter's work gets merged we can use that since the network guys
are in favor of that approach and are not going to add mempools
everywhere.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] Switch fdomain to the pci_get API
Alan Cox [Mon, 16 Oct 2006 15:55:46 +0000 (16:55 +0100)] 
[SCSI] Switch fdomain to the pci_get API

Doesn't make the hardware hot pluggable but does ensure the driver won't
crash when another device is hot-unplugged at the wrong moment. Soon I
propose to deprecate pci_find_device() and some of its friends.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] add can_queue to host parameters
James Bottomley [Mon, 16 Oct 2006 15:06:27 +0000 (10:06 -0500)] 
[SCSI] add can_queue to host parameters

Debugging TCQ issues has shown me this is a very useful parameter to be
able to view.  Add it to he host class parameters.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] megaraid_{mm,mbox}: 64-bit DMA capability fix
Andrey Mirkin [Mon, 16 Oct 2006 08:08:43 +0000 (12:08 +0400)] 
[SCSI] megaraid_{mm,mbox}: 64-bit DMA capability fix

It is known that 2 LSI Logic MegaRAID SATA RAID Controllers (150-4 and
150-6) don't support 64-bit DMA. Unfortunately currently this check is
wrong and driver sets 64-bit DMA mode for these devices.

Signed-off-by: Andrey Mirkin <amirkin@sw.ru>
Acked-by: "Ju, Seokmann" <Seokmann.Ju@lsi.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] aic94xx: Supermicro motherboards support
Sergey Kononenko [Sun, 15 Oct 2006 00:12:08 +0000 (03:12 +0300)] 
[SCSI] aic94xx: Supermicro motherboards support

Add PCI id. Plus correct for possibly missing resistor that can cause
FLASHEX to have the wrong value.

Signed-off-by: Sergey Kononenko <sergk@sergk.org.ua>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] ibmvscsi: correctly reenable CRQ
Santiago Leon [Fri, 13 Oct 2006 15:22:50 +0000 (10:22 -0500)] 
[SCSI] ibmvscsi: correctly reenable CRQ

The "ibmvscsi: treat busy and error conditions separately" patch
submitted by Dave Boutcher back in June incorrectly reenables the CRQ.
The broken logic causes the adapter to get disabled if the CRQ
connection happens to close temporarily.  This patch "fixes that
obviously wrong logic check" (Dave's words).

Signed-off-by: Santiago Leon <santil@us.ibm.com>
Signed-off-by: David Boutcher <sleddog@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] qla2xxx: Update version number to 8.01.07-k3.
Andrew Vasquez [Fri, 13 Oct 2006 16:33:40 +0000 (09:33 -0700)] 
[SCSI] qla2xxx: Update version number to 8.01.07-k3.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] qla2xxx: Correct QUEUE_FULL handling.
Andrew Vasquez [Fri, 13 Oct 2006 16:33:39 +0000 (09:33 -0700)] 
[SCSI] qla2xxx: Correct QUEUE_FULL handling.

- Drop queue-depths across all luns for a given fcport
  during TASK_SET_FULL statuses.
- Ramp-up I/Os after throttling.
- Consolidate completion-status handling of CS_QUEUE_FULL with
  CS_COMPLETE as ISP24xx firmware no longer reports
  CS_QUEUE_FULL.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] qla2xxx: Workaround D3 power-management issues.
Andrew Vasquez [Fri, 13 Oct 2006 16:33:38 +0000 (09:33 -0700)] 
[SCSI] qla2xxx: Workaround D3 power-management issues.

Early ISP2432 parts have a known hardware issue when coming
out of a D3 hot state.  This issue can result in a hung PCIe
link.  Recent firmwares contain a workaround whereby the
stop-firmware mailbox command prevents the ISP from entering
the D3 hot state.

In order to ensure that the workaround succeeded the driver
must verify that the stop-firmware mailbox command completes
successfully.  In the event of a failure, the driver
attempts a shutdown-retry after resetting the ISP and
re-executing firmware.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] qla2xxx: Check return value of sysfs_create_bin_file() usage.
Andrew Vasquez [Fri, 13 Oct 2006 16:33:37 +0000 (09:33 -0700)] 
[SCSI] qla2xxx: Check return value of sysfs_create_bin_file() usage.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] zfcp: initialize scsi_host_template.max_sectors with appropriate value
Swen Schillig [Thu, 12 Oct 2006 09:43:44 +0000 (11:43 +0200)] 
[SCSI] zfcp: initialize scsi_host_template.max_sectors with appropriate value

Define ZFCP_MAX_SECTORS and initialize scsi_host_template.max_sectors
with appropriate value.

Signed-off-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] fc4: Conversion to struct scsi_cmnd in fc4
Henrik Kretzschmar [Tue, 10 Oct 2006 21:41:42 +0000 (14:41 -0700)] 
[SCSI] fc4: Conversion to struct scsi_cmnd in fc4

Changes the obsolete Scsi_Cmnd to struct scsi_cmnd in the Fibre Channel
driver (fc4).

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] convert ninja driver to struct scsi_cmnd
Henrik Kretzschmar [Tue, 10 Oct 2006 21:41:41 +0000 (14:41 -0700)] 
[SCSI] convert ninja driver to struct scsi_cmnd

Changes the obsolete typedefd Scsi_Cmnd to struct scsi_cmnd in
the ninja scsi pcmcia driver.

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] Scsi_Cmnd conversion in psi240i driver
Henrik Kretzschmar [Tue, 10 Oct 2006 21:41:45 +0000 (14:41 -0700)] 
[SCSI] Scsi_Cmnd conversion in psi240i driver

Changes the obsolete Scsi_Cmnd to struct scsi_cmnd in psi240i-driver.

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] lpfc: check before dereference in lpfc_ct.c
Eric Sesterhenn [Tue, 10 Oct 2006 21:41:43 +0000 (14:41 -0700)] 
[SCSI] lpfc: check before dereference in lpfc_ct.c

If we fail to allocate mp->virt during the first while loop iteration,
mlist is still uninitialized, therefore we should check if before
dereferencing.

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Acked-by: James Smart <James.Smart@Emulex.Com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] scsi_lib.c: use BUILD_BUG_ON
Alexey Dobriyan [Sun, 8 Oct 2006 11:55:55 +0000 (15:55 +0400)] 
[SCSI] scsi_lib.c: use BUILD_BUG_ON

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] fix typo in previous Scsi_Cmnd convertion in aic7xxx_old.c
Henne [Mon, 9 Oct 2006 13:38:34 +0000 (15:38 +0200)] 
[SCSI] fix typo in previous Scsi_Cmnd convertion in aic7xxx_old.c

Fixes a typo in the aic7xxx_old.c.

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] mptfc: stall eh handlers if resetting while rport blocked
Michael Reed [Fri, 6 Oct 2006 20:39:25 +0000 (15:39 -0500)] 
[SCSI] mptfc: stall eh handlers if resetting while rport blocked

Thanks to James Smart for the inspiration.

Stall error handler if attempting recovery while an rport is blocked.
This avoids device offline scenarios due to errors in the error handler.
Also verify that VirtDevice is available before issuing scsi command.
VirtDevice is removed when fc transport removes a target.

See James Smart's patch of 08/17/2006 for greater detail.

http://marc.theaimsgroup.com/?l=linux-scsi&m=115583213624803&w=2

Also bump version number per Eric's request.

Signed-off-by: Michael Reed <mdr@sgi.com>
Acked-by: Eric Moore <eric.moore@lsil.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] Maintain module-parameter name consistency with qla2xxx/qla4xxx.
Andrew Vasquez [Fri, 6 Oct 2006 16:54:59 +0000 (09:54 -0700)] 
[SCSI] Maintain module-parameter name consistency with qla2xxx/qla4xxx.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] aic7xxx: Adjust .max_sectors
Hannes Reinecke [Fri, 6 Oct 2006 07:22:41 +0000 (09:22 +0200)] 
[SCSI] aic7xxx: Adjust .max_sectors

According to the adaptec sources aic7xxx / aic79xx really can do
4MB transfers. So we should adjust .max_sectors.

Signed-off-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] qla4xxx: fix double printk on load
Doug Maxey [Fri, 6 Oct 2006 04:50:07 +0000 (23:50 -0500)] 
[SCSI] qla4xxx: fix double printk on load

There is a dup printk at the tail of qla4xxx_module_init().  Remove the
first instance as it's before the complete success of the function.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] tmscsim: set max_sectors
Guennadi Liakhovetski [Thu, 5 Oct 2006 22:11:17 +0000 (00:11 +0200)] 
[SCSI] tmscsim: set max_sectors

AM53C974A's Start Transfer Counter register has 24 bits, thus
maximum transfer length is 16MiB. But the maximum I can test
is 8MiB, so use that until somebody tests 16MiB.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] st: Fixup -ENOMEDIUM
Kai Makisara [Thu, 5 Oct 2006 19:59:46 +0000 (22:59 +0300)] 
[SCSI] st: Fixup -ENOMEDIUM

Based on the original patch from Hannes Reinecke <hare@suse.de>

Fix st_open() to return -ENOMEDIUM instead of -EIO if no medium is
found.

Signed-off-by: Kai Makisara <kai.makisara@kolumbus.fi>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] Scsi_Cmnd conversion in qlogicfas408 driver
Henne [Tue, 3 Oct 2006 19:31:14 +0000 (21:31 +0200)] 
[SCSI] Scsi_Cmnd conversion in qlogicfas408 driver

Change obsolete Scsi_Cmnd to struct scsi_cmnd in the Qlocic FAS408 driver.

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
rejections fixed and
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years ago[SCSI] Scsi_Cmnd convertion in sun3-driver
Henne [Tue, 3 Oct 2006 17:51:59 +0000 (19:51 +0200)] 
[SCSI] Scsi_Cmnd convertion in sun3-driver

Change the obsolete Scsi_Cmnd to struct scsi_cmnd in the sun3-driver.

Signed-off-by: Henrik Kretzschmar <henne@nachtwindheim.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
18 years agoRemove unnecessary check in drivers/video/intelfb/intelfbhw.c
Eric Sesterhenn [Fri, 20 Oct 2006 21:35:59 +0000 (14:35 -0700)] 
Remove unnecessary check in drivers/video/intelfb/intelfbhw.c

All callers and the function itself dereference dinfo, so we can remove the
check.  (coverity id #1371)

Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Dave Airlie <airlied@linux.ie>
18 years agointel fb: switch to pci_get API
Alan Cox [Fri, 20 Oct 2006 21:36:00 +0000 (14:36 -0700)] 
intel fb: switch to pci_get API

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Dave Airlie <airlied@linux.ie>
18 years agodrm: ioremap balanced with iounmap for drivers/char/drm
Amol Lad [Wed, 25 Oct 2006 16:55:34 +0000 (09:55 -0700)] 
drm: ioremap balanced with iounmap for drivers/char/drm

ioremap must be balanced by an iounmap and failing to do so can result
in a memory leak.

Tested (compilation only) to make sure the files are compiling without
any warning/error due to new changes

Signed-off-by: Amol Lad <amol@verismonetworks.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
18 years agodrm: fix error returns, sysfs error handling
Jeff Garzik [Tue, 10 Oct 2006 21:23:37 +0000 (14:23 -0700)] 
drm: fix error returns, sysfs error handling

- callers of drm_sysfs_create() and drm_sysfs_device_add() looked for
  errors using IS_ERR(), but the functions themselves only ever returned
  NULL on error.  Fixed.

- unwind from, and propagate sysfs errors

Signed-off-by: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Dave Airlie <airlied@linux.ie>
18 years agofix return code in error case.
Dave Jones [Wed, 18 Oct 2006 04:26:39 +0000 (00:26 -0400)] 
fix return code in error case.

The other failure returns in this function are negative, so make
this one do the same.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: Dave Airlie <airlied@linux.ie>
18 years ago[CRYPTO] users: Select ECB/CBC where needed
Patrick McHardy [Wed, 25 Oct 2006 06:49:36 +0000 (16:49 +1000)] 
[CRYPTO] users: Select ECB/CBC where needed

CRYPTO_MANAGER is selected automatically by CONFIG_ECB and CONFIG_CBC.

config CRYPTO_ECB
        tristate "ECB support"
        select CRYPTO_BLKCIPHER
        select CRYPTO_MANAGER

I've added CONFIG_ECB to the ones you mentioned and CONFIG_CBC to
gssapi.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
18 years agoMerge master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6
Linus Torvalds [Wed, 25 Oct 2006 05:01:33 +0000 (22:01 -0700)] 
Merge /pub/scm/linux/kernel/git/davem/net-2.6

* master.kernel.org:/pub/scm/linux/kernel/git/davem/net-2.6:
  [DCCP]: Update documentation references.
  [ATM] horizon: read_bia() needs to be __devinit
  [NETFILTER]: Fix ip6_tables extension header bypass bug
  [NETFILTER]: Fix ip6_tables protocol bypass bug
  [XFRM]: Fix xfrm_state accounting
  [IPV4] ipconfig: fix RARP ic_servaddr breakage

18 years ago[PATCH] missing includes of io.h
Al Viro [Tue, 24 Oct 2006 10:17:06 +0000 (11:17 +0100)] 
[PATCH] missing includes of io.h

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] missing include of dma-mapping.h
Al Viro [Tue, 24 Oct 2006 10:17:37 +0000 (11:17 +0100)] 
[PATCH] missing include of dma-mapping.h

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] IOC4 should depend on PCI
Al Viro [Tue, 24 Oct 2006 10:16:29 +0000 (11:16 +0100)] 
[PATCH] IOC4 should depend on PCI

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] uml: mconsole fixes
Al Viro [Tue, 24 Oct 2006 10:15:29 +0000 (11:15 +0100)] 
[PATCH] uml: mconsole fixes

 * when we have stop/sysrq/go, we get pt_regs of whatever executes
   mc_work_proc().  Would be better to see what we had at the time of
   interrupt that got us stop.

 * stop/stop/stop.....  will give stack overflow.  Shouldn't allow stop
   from mconsole_stop().

 * stop/stop/go leaves us inside mconsole_stop() with
os_set_fd_block(req->originating_fd, 0);
reactivate_fd(req->originating_fd, MCONSOLE_IRQ);
   just done by nested mconsole_stop().  Ditto.

 * once we'd seen stop, there's a period when INTR commands are executed
   out of order (as they should; we might have the things stuck badly
   enough to never reach mconsole_stop(), but still not badly enough to
   block mconsole_interrupt(); in that situation we _want_ things like
   "cad" to be executed immediately).  Once we enter monsole_stop(), all
   INTR commands will be executed in order, mixed with PROC ones.  We'd
   better let user see that such change of behaviour has happened.
   (Suggested by lennert).

 * stack footprint of monsole_interrupt() is an atrocity; AFAICS we can
   safely make struct mc_request req; static in function there.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[PATCH] Merge headphone and speaker volume controls for Panasonic R4 laptop
Keith Packard [Tue, 24 Oct 2006 20:34:11 +0000 (13:34 -0700)] 
[PATCH] Merge headphone and speaker volume controls for Panasonic R4 laptop

Signed-off-by: Keith Packard <keithp@neko.keithp.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
18 years ago[POWERPC] Fix CHRP platforms with only 8259
Benjamin Herrenschmidt [Wed, 25 Oct 2006 03:22:27 +0000 (13:22 +1000)] 
[POWERPC] Fix CHRP platforms with only 8259

On CHRP platforms with only a 8259 controller, we should set the
default IRQ host to the 8259 driver's one for the IRQ probing
fallbacks to work in case the IRQ tree is incorrect (like on
Pegasos for example). Without this fix, we get a bunch of WARN_ON's
during boot.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] IPIC: Fix spinlock recursion in set_irq_handler
Scott Wood [Mon, 23 Oct 2006 16:35:22 +0000 (11:35 -0500)] 
[POWERPC] IPIC: Fix spinlock recursion in set_irq_handler

This causes ipic_set_irq_type to set the handler directly rather
than call set_irq_handler, which causes spinlock recursion because
the lock is already held when ipic_set_irq_type is called.

I'm also not convinced that ipic_set_irq_type should be changing the
handler at all.  There seem to be several controllers that don't and
several that do.  Those that do would break what appears to be a common
usage of calling set_irq_chip_and_handler followed by set_irq_type, if a
non-standard handler were to be used.  OTOH, irq_create_of_mapping()
doesn't set the handler, but only calls set_irq_type().

This patch gets things working in the spinlock-debugging-enabled case,
but I'm curious as to where the handler setting is ideally supposed to be
done.  I don't see any documentation on set_irq_type() that clarifies
what the semantics are supposed to be.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Fix the UCC rx/tx clock of QE
Liu Dave-r63238 [Wed, 18 Oct 2006 08:36:56 +0000 (16:36 +0800)] 
[POWERPC] Fix the UCC rx/tx clock of QE

MPC8323EMDS board ethernet interface with RMII uses the CLK16 divisor
for the rx and tx clock, but the ucc_set_qe_mux_rxtx() function doesn't
handle the CLK16 setting of the CMXUCR3 and CMXUCR4 registers.  This
fixes it.

Signed-off-by: Dave Liu <daveliu@freescale.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] cell: update defconfig
Arnd Bergmann [Tue, 24 Oct 2006 16:01:43 +0000 (18:01 +0200)] 
[POWERPC] cell: update defconfig

===================================================================

Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] spufs: fix another off-by-one bug in spufs_mbox_read
Arnd Bergmann [Tue, 24 Oct 2006 16:01:42 +0000 (18:01 +0200)] 
[POWERPC] spufs: fix another off-by-one bug in spufs_mbox_read

Currently, spufs_mbox_read transfers more bytes than requested on a
read.  If you ask for four bytes, you get eight.  This fixes it to
transfer the largest multiple of four bytes that is less than or equal
to the number you asked for.

Note: one nasty property of this file in spufs is that you can only
read multiples of four bytes in the first place, since there is no way
to atomically put back a few bytes into the hardware register.  Thus,
reading less than four bytes returns -EINVAL.  Asking for more than
four returns the largest possible multiple of four.

Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] spufs: fix signal2 file to report signal2
Dwayne Grant Mcconnell [Tue, 24 Oct 2006 16:27:30 +0000 (18:27 +0200)] 
[POWERPC] spufs: fix signal2 file to report signal2

This fixes the /signal2 file to actually give signal2 data.

Signed-off-by: Dwayne Grant Mcconnell <decimal@us.ibm.com>
Signed-off-by: Arnd Bergmann <arnd.bergmann@de.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Fix device_is_compatible() const warning
Benjamin Herrenschmidt [Fri, 20 Oct 2006 01:49:54 +0000 (11:49 +1000)] 
[POWERPC] Fix device_is_compatible() const warning

Fix a const'ification related warning with device_is_compatible()
and friends related to get_property() not properly having const
on it's input device node argument.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Cell timebase bug workaround
Benjamin Herrenschmidt [Fri, 20 Oct 2006 04:37:05 +0000 (14:37 +1000)] 
[POWERPC] Cell timebase bug workaround

The Cell CPU timebase has an erratum. When reading the entire 64 bits
of the timebase with one mftb instruction, there is a handful of cycles
window during which one might read a value with the low order 32 bits
already reset to 0x00000000 but the high order bits not yet incremeted
by one. This fixes it by reading the timebase again until the low order
32 bits is no longer 0. That might introduce occasional latencies if
hitting mftb just at the wrong time, but no more than 70ns on a cell
blade, and that was considered acceptable.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Support feature fixups in modules
Benjamin Herrenschmidt [Fri, 20 Oct 2006 01:47:19 +0000 (11:47 +1000)] 
[POWERPC] Support feature fixups in modules

This patch adds support for feature fixups in modules. This involves
adding support for R_PPC64_REL64 relocs to the 64 bits module loader.
It also modifies modpost.c to ignore the powerpc fixup sections (or it
would warn when used in .init.text).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Support feature fixups in vdso's
Benjamin Herrenschmidt [Fri, 20 Oct 2006 01:47:18 +0000 (11:47 +1000)] 
[POWERPC] Support feature fixups in vdso's

This patch reworks the feature fixup mecanism so vdso's can be fixed up.
The main issue was that the construct:

        .long   label  (or .llong on 64 bits)

will not work in the case of a shared library like the vdso. It will
generate an empty placeholder in the fixup table along with a reloc,
which is not something we can deal with in the vdso.

The idea here (thanks Alan Modra !) is to instead use something like:

1:
        .long   label - 1b

That is, the feature fixup tables no longer contain addresses of bits of
code to patch, but offsets of such code from the fixup table entry
itself. That is properly resolved by ld when building the .so's. I've
modified the fixup mecanism generically to use that method for the rest
of the kernel as well.

Another trick is that the 32 bits vDSO included in the 64 bits kernel
need to have a table in the 64 bits format. However, gas does not
support 32 bits code with a statement of the form:

        .llong  label - 1b  (Or even just .llong label)

That is, it cannot emit the right fixup/relocation for the linker to use
to assign a 32 bits address to an .llong field. Thus, in the specific
case of the 32 bits vdso built as part of the 64 bits kernel, we are
using a modified macro that generates:

        .long   0xffffffff
        .llong  label - 1b

Note that is assumes that the value is negative which is enforced by
the .lds (those offsets are always negative as the .text is always
before the fixup table and gas doesn't support emiting the reloc the
other way around).

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Support nested cpu feature sections
Benjamin Herrenschmidt [Fri, 20 Oct 2006 01:47:16 +0000 (11:47 +1000)] 
[POWERPC] Support nested cpu feature sections

This patch adds some macros that can be used with an explicit label in
order to nest cpu features. This should be used very careful but is
necessary for the upcoming cell TB fixup.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Consolidate feature fixup code
Benjamin Herrenschmidt [Tue, 24 Oct 2006 06:42:40 +0000 (16:42 +1000)] 
[POWERPC] Consolidate feature fixup code

There are currently two versions of the functions for applying the
feature fixups, one for CPU features and one for firmware features. In
addition, they are both in assembly and with separate implementations
for 32 and 64 bits. identify_cpu() is also implemented in assembly and
separately for 32 and 64 bits.

This patch replaces them with a pair of C functions. The call sites are
slightly moved on ppc64 as well to be called from C instead of from
assembly, though it's a very small change, and thus shouldn't cause any
problem.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Olof Johansson <olof@lixom.net>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[POWERPC] Fix hang in start_ldr if _end or _edata is unaligned
Olaf Hering [Fri, 20 Oct 2006 13:57:15 +0000 (15:57 +0200)] 
[POWERPC] Fix hang in start_ldr if _end or _edata is unaligned

Quick fix for lack of memset(__bss_start, 0, _end-__bss_start) in
load_kernel().  If edata is unaligned, the loop will overwrite all
memory because r3 and r4 will never be equal.

Signed-off-by: Olaf Hering <olaf@aepfle.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
18 years ago[DCCP]: Update documentation references.
Gerrit Renker [Tue, 24 Oct 2006 23:17:51 +0000 (16:17 -0700)] 
[DCCP]: Update documentation references.

Updates the references to spec documents throughout the code, taking into
account that

* the DCCP, CCID 2, and CCID 3 drafts all became RFCs in March this year

* RFC 1063 was obsoleted by RFC 1191

* draft-ietf-tcpimpl-pmtud-0x.txt was published as an Informational
  RFC, RFC 2923 on 2000-09-22.

All references verified.

Signed-off-by: Gerrit Renker <gerrit@erg.abdn.ac.uk>
Signed-off-by: Arnaldo Carvalho de Melo <acme@mandriva.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[ATM] horizon: read_bia() needs to be __devinit
David S. Miller [Tue, 24 Oct 2006 23:16:39 +0000 (16:16 -0700)] 
[ATM] horizon: read_bia() needs to be __devinit

Thanks to Randy Dunlap.

Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NETFILTER]: Fix ip6_tables extension header bypass bug
Patrick McHardy [Tue, 24 Oct 2006 23:15:10 +0000 (16:15 -0700)] 
[NETFILTER]: Fix ip6_tables extension header bypass bug

As reported by Mark Dowd <Mark_Dowd@McAfee.com>, ip6_tables is susceptible
to a fragmentation attack causing false negatives on extension header matches.

When extension headers occur in the non-first fragment after the fragment
header (possibly with an incorrect nexthdr value in the fragment header)
a rule looking for this extension header will never match.

Drop fragments that are at offset 0 and don't contain the final protocol
header regardless of the ruleset, since this should not happen normally.
Since all extension headers are before the protocol header this makes sure
an extension header is either not present or in the first fragment, where
we can properly parse it.

With help from Yasuyuki KOZAKAI <yasuyuki.kozakai@toshiba.co.jp>.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[NETFILTER]: Fix ip6_tables protocol bypass bug
Patrick McHardy [Tue, 24 Oct 2006 23:14:04 +0000 (16:14 -0700)] 
[NETFILTER]: Fix ip6_tables protocol bypass bug

As reported by Mark Dowd <Mark_Dowd@McAfee.com>, ip6_tables is susceptible
to a fragmentation attack causing false negatives on protocol matches.

When the protocol header doesn't follow the fragment header immediately,
the fragment header contains the protocol number of the next extension
header. When the extension header and the protocol header are sent in
a second fragment a rule like "ip6tables .. -p udp -j DROP" will never
match.

Drop fragments that are at offset 0 and don't contain the final protocol
header regardless of the ruleset, since this should not happen normally.

With help from Yasuyuki KOZAKAI <yasuyuki.kozakai@toshiba.co.jp>.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[PATCH] x86-64: Only look at per_cpu data for online cpus.
Eric W. Biederman [Tue, 24 Oct 2006 23:00:23 +0000 (01:00 +0200)] 
[PATCH] x86-64: Only look at per_cpu data for online cpus.

When I generalized __assign_irq_vector I failed to pay attention
to what happens when you access a per cpu data structure for
a cpu that is not online.   It is an undefined case making any
code that does it have undefined behavior as well.

The code still needs to be able to allocate a vector across cpus
that are not online to properly handle combinations like lowest
priority interrupt delivery and cpu_hotplug.  Not that we can do
that today but the infrastructure shouldn't prevent it.

So this patch updates the places where we touch per cpu data
to only touch online cpus, it makes cpu vector allocation
an atomic operation with respect to cpu hotplug, and it updates
the cpu start code to properly initialize vector_irq so we
don't have inconsistencies.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andi Kleen <ak@suse.de>
18 years ago[PATCH] x86-64: Simplify the vector allocator.
Eric W. Biederman [Tue, 24 Oct 2006 23:00:22 +0000 (01:00 +0200)] 
[PATCH] x86-64: Simplify the vector allocator.

There is no reason to remember a per cpu position of which vector
to try.  Keeping a global position is simpler and more likely to
result in a global vector allocation even if I don't need or require
it.  For level triggered interrupts this means we are less likely to
acknowledge another cpus irq, and cause the level triggered irq to
harmlessly refire.

This simplification makes it easier to only access data structures
of  online cpus, by having fewer special cases to deal with.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Andi Kleen <ak@suse.de>
18 years ago[XFRM]: Fix xfrm_state accounting
Patrick McHardy [Tue, 24 Oct 2006 22:34:00 +0000 (15:34 -0700)] 
[XFRM]: Fix xfrm_state accounting

xfrm_state_num needs to be increased for XFRM_STATE_ACQ states created
by xfrm_state_find() to prevent the counter from going negative when
the state is destroyed.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years ago[IPV4] ipconfig: fix RARP ic_servaddr breakage
Al Viro [Tue, 24 Oct 2006 22:18:36 +0000 (15:18 -0700)] 
[IPV4] ipconfig: fix RARP ic_servaddr breakage

memcpy 4 bytes to address of auto unsigned long variable followed
by comparison with u32 is a bloody bad idea.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
18 years agoe100: account for closed interface when shutting down
Auke Kok [Tue, 24 Oct 2006 21:49:44 +0000 (14:49 -0700)] 
e100: account for closed interface when shutting down

Account for the interface being closed before disabling polling
on a device, to fix shutdown on some systems that explcitly close
the netdevice before calling shutdown.

Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
18 years agoe1000: Increment version to 7.2.9-k4
Auke Kok [Tue, 24 Oct 2006 21:46:09 +0000 (14:46 -0700)] 
e1000: Increment version to 7.2.9-k4

Significant fixes -> increment driver version.

Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
18 years agoe1000: FIX: move length adjustment due to crc stripping disabled.
Jesse Brandeburg [Tue, 24 Oct 2006 21:46:06 +0000 (14:46 -0700)] 
e1000: FIX: move length adjustment due to crc stripping disabled.

Move the length (rx_bytes counter) adjustment of 4 bytes down to after the
TBI_ACCEPT workaround.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
18 years agoe1000: FIX: Don't limit descriptor size to 4kb for PCI-E adapters
Jesse Brandeburg [Tue, 24 Oct 2006 21:46:04 +0000 (14:46 -0700)] 
e1000: FIX: Don't limit descriptor size to 4kb for PCI-E adapters

82571 and newer chispets don't need to limit desc. length to 4kb and can
handle 8kb sizes.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
18 years agoe1000: FIX: Disable Packet Split for non jumbo frames
Jesse Brandeburg [Tue, 24 Oct 2006 21:46:01 +0000 (14:46 -0700)] 
e1000: FIX: Disable Packet Split for non jumbo frames

Allocations using alloc_page are taking too long for normal MTU, so
use LPE only for jumbo frames.

Signed-off-bu: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
18 years agoe1000: FIX: fix wrong txdctl threshold bitmasks
Bruce Allan [Tue, 24 Oct 2006 21:45:58 +0000 (14:45 -0700)] 
e1000: FIX: fix wrong txdctl threshold bitmasks

Threshold bitmasks for prefetch, host and writeback were clearing
bits that they were not supposed to. The leftmost 2 bits in the byte
for each threshold are reserved.

Signed-off-by: Bruce Allan <bruce.w.allan@intel.com>
Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com>
This page took 0.045421 seconds and 5 git commands to generate.