safe/jmp/linux-2.6
19 years ago[PATCH] 2.6.12-rc1-mm3 Fix ver_linux script for no udev utils.
Steven Cole [Tue, 19 Apr 2005 04:57:33 +0000 (21:57 -0700)]
[PATCH] 2.6.12-rc1-mm3 Fix ver_linux script for no udev utils.

Without the attached patch, the ver_linux script gives
the following if udev utils are not present.

./scripts/ver_linux: line 90: udevinfo: command not found

The patch causes ver_linux to be silent in the case of
no udevinfo command.

Signed-off-by: Steven Cole <elenstev@mesatop.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
19 years ago[PATCH] export platform_add_devices
Robert Schwebel [Tue, 19 Apr 2005 04:57:32 +0000 (21:57 -0700)]
[PATCH] export platform_add_devices

platform_add_devices can be used from within modules, so it should be
exported.  This can for example happen if you have hotpluggable firmware in
an FPGA on a system on chip processor; in our case the FPGA is probed for
devices and the FPGA base code registers the devices it has found with the
kernel.

(akpm: I think this is reasonable from a licensing POV: it's unlikely that
anyone would be interested in merging such specialised modules into mainline,
and it's a GPL export).

Signed-off-by: Robert Schwebel <r.schwebel@pengutronix.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
19 years ago[PATCH] sysfs: add sysfs_chmod_file()
Kay Sievers [Tue, 19 Apr 2005 04:57:32 +0000 (21:57 -0700)]
[PATCH] sysfs: add sysfs_chmod_file()

sysfs: allow changing the permissions for already created attributes

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
19 years ago[PATCH] add TIMEOUT to firmware_class hotplug event
kay.sievers@vrfy.org [Tue, 19 Apr 2005 04:57:31 +0000 (21:57 -0700)]
[PATCH] add TIMEOUT to firmware_class hotplug event

On Tue, 2005-03-15 at 09:25 +0100, Hannes Reinecke wrote:
> The current implementation of the firmware class breaks a fundamental
> assumption in udevd: that the physical device can be initialised fully
> prior to executing the next event for that device.

Here we add a TIMEOUT value to the hotplug environment of the firmware
requesting event. I will adapt udevd not to wait for anything else, if
it finds a TIMEOUT key.

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
19 years ago[PATCH] kref: add link to original documentation to the kref documentation.
gregkh@suse.de [Tue, 19 Apr 2005 04:57:30 +0000 (21:57 -0700)]
[PATCH] kref: add link to original documentation to the kref documentation.

Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
19 years ago[PATCH] kref: add documentation
minyard@acm.org [Tue, 19 Apr 2005 04:57:30 +0000 (21:57 -0700)]
[PATCH] kref: add documentation

Add some documentation for krefs.

Signed-off-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
19 years agofully merge up to scsi-misc-2.6
James Bottomley [Tue, 19 Apr 2005 01:55:09 +0000 (20:55 -0500)]
fully merge up to scsi-misc-2.6

19 years agoMerge with master.kernel.org:/home/rmk/linux-2.6-rmk.git
Linus Torvalds [Mon, 18 Apr 2005 23:25:10 +0000 (16:25 -0700)]
Merge ... kernel.org:/home/rmk/linux-2.6-rmk.git

This adds the missing arch/arm/lib/bitops.h file.

19 years ago[PATCH] sparc64: Fix stat
David S. Miller [Mon, 18 Apr 2005 22:13:15 +0000 (15:13 -0700)]
[PATCH] sparc64: Fix stat

Like Alpha, sparc64's struct stat was defined before we had the
nanosecond et al.  fields added.  So like Alpha I have to cons up a
struct stat64 to get this stuff.  I'll work on the glibc bits soon.

Also, we were forgetting to fill in the nanosecond fields in the sparc
compat stat64 syscalls.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ARM: Add missing new file for bitops patch
Russell King [Mon, 18 Apr 2005 21:50:01 +0000 (22:50 +0100)]
[PATCH] ARM: Add missing new file for bitops patch

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
19 years agoMerge SCSI tree from James Bottomley.
Linus Torvalds [Mon, 18 Apr 2005 21:25:40 +0000 (14:25 -0700)]
Merge SCSI tree from James Bottomley.

Done with "git-pull-script rsync://www.parisc-linux.org/~jejb/scsi-rc-fixes-2.6.git"
together with an automated content merge.

19 years agoaic7xxx: convert to SPI transport class Domain Validation
<jejb@titanic.il.steeleye.com> [Sun, 17 Apr 2005 23:03:20 +0000 (18:03 -0500)]
aic7xxx: convert to SPI transport class Domain Validation

Now that we export all the parameters, this is easy to do.
It also means that we can dump about 2000 lines of code that
were dedicated to doing this internally.

Additionally, this removes all the aic7xxx driver abuse
of SCSI timers which were embedded in the DV routines.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agoaic7xxx: add support for the SPI transport class
<jejb@titanic.il.steeleye.com> [Sun, 17 Apr 2005 21:59:33 +0000 (16:59 -0500)]
aic7xxx: add support for the SPI transport class

This is just a simplistic patch to export all of the
aic7xxx internal transport parameters via the SPI
transport class.  It doesn't actually alter the way the
driver works at all.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agoqla2xxx: fix compiler warning in qla_attr.c
<jejb@titanic.il.steeleye.com> [Sun, 17 Apr 2005 21:32:42 +0000 (16:32 -0500)]
qla2xxx: fix compiler warning in qla_attr.c

  CC [M]  drivers/scsi/qla2xxx/qla_attr.o
drivers/scsi/qla2xxx/qla_attr.c: In function `qla2x00_sysfs_write_fw_dump':
drivers/scsi/qla2xxx/qla_attr.c:65: warning: implicit declaration of function `vfree'
drivers/scsi/qla2xxx/qla_attr.c:83: warning: implicit declaration of function `vmalloc'
drivers/scsi/qla2xxx/qla_attr.c:83: warning: cast to pointer from integer of different size

Also remove spurious inclusion of linux/version.h

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agolpfc: add Emulex FC driver version 8.0.28
<jejb@titanic.il.steeleye.com> [Sun, 17 Apr 2005 21:05:31 +0000 (16:05 -0500)]
lpfc: add Emulex FC driver version 8.0.28

From:  James.Smart@Emulex.Com

Modified for kernel import and
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] drivers/scsi/gdth.c: cleanups
<bunk@stusta.de> [Sun, 17 Apr 2005 20:28:39 +0000 (15:28 -0500)]
[PATCH] drivers/scsi/gdth.c: cleanups

This patch contains the following cleanups:
- make some needlessly global functions static
- remove one more kernel 2.2 #ifdef

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] remove old scsi data direction macros
<hch@lst.de> [Sun, 17 Apr 2005 20:26:13 +0000 (15:26 -0500)]
[PATCH] remove old scsi data direction macros

these have been wrappers for the generic dma direction bits since 2.5.x.
This patch converts the few remaining drivers and removes the macros.

Arjan noticed there's some hunk in here that shouldn't.  Updated patch
below:

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] qla2xxx: update version to 8.00.02b5-k
<andrew.vasquez@qlogic.com> [Sun, 17 Apr 2005 20:16:47 +0000 (15:16 -0500)]
[PATCH] qla2xxx: update version to 8.00.02b5-k

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] qla2xxx: remove sale revision notes file
<andrew.vasquez@qlogic.com> [Sun, 17 Apr 2005 20:14:51 +0000 (15:14 -0500)]
[PATCH] qla2xxx: remove sale revision notes file

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] qla2xxx: remove /proc interface
<andrew.vasquez@qlogic.com> [Sun, 17 Apr 2005 20:13:01 +0000 (15:13 -0500)]
[PATCH] qla2xxx: remove /proc interface

Remove /proc support.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] qla2xxx: cleanup DMA mappings...
<andrew.vasquez@qlogic.com> [Sun, 17 Apr 2005 20:10:41 +0000 (15:10 -0500)]
[PATCH] qla2xxx: cleanup DMA mappings...

  Don't use cmd->request->nr_hw_segments as it may not be initialized
  (SG_IO in particular bypasses anything that initializes this and just
  uses scsi_do_req to insert a scsi_request directly on the head of the
  queue) and a bogus value here can trip up the checks to make sure that
  the number of segments will fit in the queue ring buffer, resulting in
  commands that are never completed.

  Fix up several issues with PCI DMA mapping and failure to check return
  values on the mappings.

  Make the check for space in the ring buffer happen after the DMA mapping
  is done since any checks done before the mapping has taken place are
  bogus.

  Doug Ledford <dledford@redhat.com>.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] qla2xxx: remove lun discovery codes...
<andrew.vasquez@qlogic.com> [Sun, 17 Apr 2005 20:06:53 +0000 (15:06 -0500)]
[PATCH] qla2xxx: remove lun discovery codes...

Remove internal lun discovery routines and support
structures.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] qla2xxx: add remote port codes...
<andrew.vasquez@qlogic.com> [Sun, 17 Apr 2005 20:04:54 +0000 (15:04 -0500)]
[PATCH] qla2xxx: add remote port codes...

Add initial support for FC remote port infrastructure.

     o Use fc_remote_port...() registration and block/unlock
       functions.
     o Consolidate 'attribute' (fc-remote/sysfs) helpers into
       new qla_attr.c file.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] qla2xxx: remove internal queuing...
<andrew.vasquez@qlogic.com> [Sun, 17 Apr 2005 20:02:26 +0000 (15:02 -0500)]
[PATCH] qla2xxx: remove internal queuing...

Remove internal command queuing from the driver.  As is, this
driver cannot tolerate cable-pulls as I/Os will begin to fail
by the upper layers.

     o Should be used in conjuction with the
       11-fc_rport_adds_2.diff patch.
     o Removes qla_listops.h file -- no longer needed.

Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agomerge by hand (scsi_device.h)
James Bottomley [Mon, 18 Apr 2005 18:45:00 +0000 (13:45 -0500)]
merge by hand (scsi_device.h)

19 years ago[PATCH] sched: fix signed comparisons of long long
Ingo Molnar [Mon, 18 Apr 2005 17:58:36 +0000 (10:58 -0700)]
[PATCH] sched: fix signed comparisons of long long

Signed-off-by: Jesper Juhl <juhl-lkml@dif.dk>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] Add 32-bit compatibility for NFSv4 mount
David Howells [Mon, 18 Apr 2005 17:54:51 +0000 (10:54 -0700)]
[PATCH] Add 32-bit compatibility for NFSv4 mount

This adds 32-bit compatibility for mounting an NFSv4 mount on a 64-bit
kernel (such as happens with PPC64).

The problem is that the mount data for the NFS4 mount process includes
auxilliary data pointers, probably because the NFS4 mount data may
conceivably exceed PAGE_SIZE in size - thus breaking against the hard
limit imposed by sys_mount().

Signed-Off-By: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] SELinux: fix deadlock on dcache lock
Stephen Smalley [Mon, 18 Apr 2005 17:47:35 +0000 (10:47 -0700)]
[PATCH] SELinux: fix deadlock on dcache lock

This fixes a deadlock on the dcache lock detected during testing at IBM
by moving the logging of the current executable information from the
SELinux avc_audit function to audit_log_exit (via an audit_log_task_info
helper) for processing upon syscall exit.

For consistency, the patch also removes the logging of other
task-related information from avc_audit, deferring handling to
audit_log_exit instead.

This allows simplification of the avc_audit code, allows the exe
information to be obtained more reliably, always includes the comm
information (useful for scripts), and avoids including bogus task
information for checks performed from irq or softirq.

Signed-off-by: Stephen Smalley <sds@tycho.nsa.gov>
Signed-off-by: James Morris <jmorris@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] M68k: Update defconfigs for 2.6.12-rc2
Geert Uytterhoeven [Mon, 18 Apr 2005 17:47:34 +0000 (10:47 -0700)]
[PATCH] M68k: Update defconfigs for 2.6.12-rc2

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] M68k: Update defconfigs for 2.6.11
Geert Uytterhoeven [Mon, 18 Apr 2005 17:47:33 +0000 (10:47 -0700)]
[PATCH] M68k: Update defconfigs for 2.6.11

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] scsi: scsi_send_eh_cmnd() cleanup
<htejun@gmail.com> [Sun, 17 Apr 2005 20:00:23 +0000 (15:00 -0500)]
[PATCH] scsi: scsi_send_eh_cmnd() cleanup

This patch makes scsi_send_eh_cmnd() use sdev and shost instead of
referencing them through scmd-> everytime.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] scsi: remove volatile from scsi data
<htejun@gmail.com> [Sun, 17 Apr 2005 19:57:29 +0000 (14:57 -0500)]
[PATCH] scsi: remove volatile from scsi data

This patch removes volatile qualifier from scsi_device->device_busy,
Scsi_Host->host_busy and ->host_failed as the volatile qualifiers
don't serve any purpose now.  While at it, convert those fields from
unsigned short to unsigned int as suggested by Christoph.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agoscsi: add DID_REQUEUE to the error handling
<jejb@titanic.il.steeleye.com> [Tue, 12 Apr 2005 22:49:09 +0000 (17:49 -0500)]
scsi: add DID_REQUEUE to the error handling

We have a DID_IMM_RETRY to require a retry at once, but we could do with
a DID_REQUEUE to instruct the mid-layer to treat this command in the
same manner as QUEUE_FULL or BUSY (i.e. halt the submission until
another command returns ... or the queue pressure builds if there are no
outstanding commands).

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agozfcp: add point-2-point support
<jejb@titanic.il.steeleye.com> [Mon, 11 Apr 2005 04:04:28 +0000 (23:04 -0500)]
zfcp: add point-2-point support

From: Andreas Herrmann <aherrman@de.ibm.com>

This patch mainly introduces support for point-2-point
topology.

From: Heiko Carstens <heiko.carstens@de.ibm.com>
From: Maxim Shchetynin <maxim@de.ibm.com>
From: Andreas Herrmann <aherrman@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] Convert i2o to compat_ioctl
<jejb@titanic.il.steeleye.com> [Mon, 11 Apr 2005 03:29:42 +0000 (22:29 -0500)]
[PATCH] Convert i2o to compat_ioctl

Also broken design in its compat handlers - CONFIG_COMPAT doesn't
mean that there should be no native ioctls.

Signed-off-by: Andi Kleen <ak@muc.de>
Reworked with comments from Markus Lidel by:
Christoph Hellwig <hch@infradead.org>

Rejections fixed and
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] kill old EH constants
<hch@lst.de> [Mon, 11 Apr 2005 03:19:25 +0000 (22:19 -0500)]
[PATCH] kill old EH constants

Fix up two drivers that incorrectly were using the old return values for
their new-style EH methods and kill off scsi_obsolete.h that defined the
constants.  The initio driver has all these constansts defined locally
and uses them internally, I'll fix that up some time later.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] scsi: remove meaningless scsi_cmnd->serial_number_at_timeout field
<htejun@gmail.com> [Sun, 3 Apr 2005 19:59:11 +0000 (14:59 -0500)]
[PATCH] scsi: remove meaningless scsi_cmnd->serial_number_at_timeout field

scsi_cmnd->serial_number_at_timeout doesn't serve any purpose
anymore.  All serial_number == serial_number_at_timeout tests
are always true in abort callbacks.  Kill the field.  Also, as
->pid always equals ->serial_number and ->serial_number
doesn't have any special meaning anymore, update comments
above ->serial_number accordingly.  Once we remove all uses of
this field from all lldd's, this field should go.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] scsi: remove unused scsi_cmnd->internal_timeout field
<htejun@gmail.com> [Sun, 3 Apr 2005 19:55:06 +0000 (14:55 -0500)]
[PATCH] scsi: remove unused scsi_cmnd->internal_timeout field

scsi_cmnd->internal_timeout field doesn't have any meaning
anymore.  Kill the field.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] remove outdated print_* functions
<hch@lst.de> [Sun, 3 Apr 2005 19:53:59 +0000 (14:53 -0500)]
[PATCH] remove outdated print_* functions

We have the scsi_print_* functions in the proper namespace for a long
time now and there weren't a lot users left.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] consolidate timeout defintions in scsi.h
<hch@lst.de> [Sun, 3 Apr 2005 19:52:44 +0000 (14:52 -0500)]
[PATCH] consolidate timeout defintions in scsi.h

Adapted from a patch in SuSE's kernel SRPM.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] x86: fix acpi compile without CONFIG_ACPI_BUS
Chris Wedgwood [Mon, 18 Apr 2005 15:01:30 +0000 (08:01 -0700)]
[PATCH] x86: fix acpi compile without CONFIG_ACPI_BUS

The recent acpi boot patch breaks for me: acpi_fadt needs CONFIG_ACPI_BUS.

Signed-off-By: Chris Wedgwood <cw@f00f.org>
Acked-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] sparc64: Fix copy_sigingo_to_user32()
Jurij Smakov [Mon, 18 Apr 2005 01:03:12 +0000 (18:03 -0700)]
[PATCH] sparc64: Fix copy_sigingo_to_user32()

The compat routine to copy over this data structure was not
handling SI_POLL correctly, breaking various fcntl() variants
in compat tasks.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] sparc64: Reduce ptrace cache flushing
David S. Miller [Mon, 18 Apr 2005 01:03:11 +0000 (18:03 -0700)]
[PATCH] sparc64: Reduce ptrace cache flushing

We were flushing the D-cache excessively for ptrace() processing
and this makes debugging threads so slow as to be totally unusable.

All process page accesses via ptrace() go via access_process_vm().
This routine, for each process page, uses get_user_pages().  That
in turn does a flush_dcache_page() on the child pages before we
copy in/out the ptrace request data.

Therefore, all we need to do after the data movement is:

1) Flush the D-cache pages if the kernel maps the page to a different
   color than userspace does.
2) If we wrote to the page, we need to flush the I-cache on older cpus.

Previously we just flushed the entire cache at the end of a ptrace()
request, and that was beyond stupid.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] sparc: Fix PTRACE_CONT bogosity
David S. Miller [Mon, 18 Apr 2005 01:03:11 +0000 (18:03 -0700)]
[PATCH] sparc: Fix PTRACE_CONT bogosity

SunOS aparently had this weird PTRACE_CONT semantic which
we copied.  If the addr argument is something other than
1, it sets the process program counter to whatever that
value is.

This is different from every other Linux architecture, which
don't do anything with the addr and data args.

This difference in particular breaks the Linux native GDB support
for fork and vfork tracing on sparc and sparc64.

There is no interest in running SunOS binaries using this weird
PTRACE_CONT behavior, so just delete it so we behave like other
platforms do.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] sparc64: use message queue compat syscalls
David S. Miller [Mon, 18 Apr 2005 01:03:10 +0000 (18:03 -0700)]
[PATCH] sparc64: use message queue compat syscalls

A couple message queue system call entries for compat tasks
were not using the necessary compat_sys_*() functions, causing
some glibc test cases to fail.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] sparc64: Do not flush dcache for ZERO_PAGE.
David S. Miller [Mon, 18 Apr 2005 01:03:09 +0000 (18:03 -0700)]
[PATCH] sparc64: Do not flush dcache for ZERO_PAGE.

This case actually can get exercised a lot during an ELF
coredump of a process which contains a lot of non-COW'd
anonymous pages.  GDB has this test case which in partiaular
creates near terabyte process full of ZERO_PAGEes.  It takes
forever to just walk through the page tables because of
all of these spurious cache flushes on sparc64.

With this change it takes only a second or so.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years agoMerge with master.kernel.org:/home/rmk/linux-2.6-rmk.git - ARM changes
Linus Torvalds [Sun, 17 Apr 2005 21:47:24 +0000 (14:47 -0700)]
Merge ... /home/rmk/linux-2.6-rmk.git - ARM changes

First ever true git merge. Let's see if it actually works.

19 years agoFix up some file mode differences due to the new git world order.
Linus Torvalds [Sun, 17 Apr 2005 20:30:46 +0000 (13:30 -0700)]
Fix up some file mode differences due to the new git world order.

19 years ago[PATCH] revert fs/char_dev.c CONFIG_BASE_FULL change
David Brownell [Sun, 17 Apr 2005 17:57:20 +0000 (10:57 -0700)]
[PATCH] revert fs/char_dev.c CONFIG_BASE_FULL change

This reverts a fs/char_dev.c patch that was merged into BK on March 3.

The problem is that it breaks things ... __register_chrdev_region() has
a block of code, commented "temporary" for over two years now, which
fails rudely during PCMCIA initialization or other register_chrdev()
calls, because it doesn't "degrade to linked list".  This keeps whole
subsystems from working.

A real fix to that "temporary" code should be possible, using some better
scheme to allocate major numbers, but it's not something I want to spend
time on just now.

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Acked-by: Matt Mackall <mpm@selenic.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] ARM: fix debug macros
Russell King [Sun, 17 Apr 2005 15:28:31 +0000 (16:28 +0100)]
[PATCH] ARM: fix debug macros

Fix debug EBSA285 and RiscPC debugging macros to detect whether the
MMU is enabled.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
19 years ago[PATCH] ARM: bitops
Russell King [Sun, 17 Apr 2005 14:51:02 +0000 (15:51 +0100)]
[PATCH] ARM: bitops

Convert ARM bitop assembly to a macro.  All bitops follow the same
format, so it's silly duplicating the code when only one or two
instructions are different.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
19 years ago[PATCH] ARM: showregs
Russell King [Sun, 17 Apr 2005 14:50:36 +0000 (15:50 +0100)]
[PATCH] ARM: showregs

Fix show_regs() to provide a backtrace.  Provide a new __show_regs()
function which implements the common subset of show_regs() and die().
Add prototypes to asm-arm/system.h

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
19 years ago[PATCH] ARM: h3600_irda_set_speed arguments
Russell King [Sun, 17 Apr 2005 14:40:46 +0000 (15:40 +0100)]
[PATCH] ARM: h3600_irda_set_speed arguments

h3600_irda_set_speed() had the wrong type for the "speed" argument.
Fix this.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
19 years ago[PATCH] ARM: footbridge rtc init
Russell King [Sun, 17 Apr 2005 14:36:55 +0000 (15:36 +0100)]
[PATCH] ARM: footbridge rtc init

The footbridge ISA RTC was being initialised before we had setup the
kernel timer.  This caused a divide by zero error when the current
time of day is set.  Resolve this by initialising the RTC after
the kernel timer has been initialised.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
19 years agoscsi: add DID_REQUEUE to the error handling
<jejb@titanic.il.steeleye.com> [Tue, 12 Apr 2005 22:49:09 +0000 (17:49 -0500)]
scsi: add DID_REQUEUE to the error handling

We have a DID_IMM_RETRY to require a retry at once, but we could do with
a DID_REQUEUE to instruct the mid-layer to treat this command in the
same manner as QUEUE_FULL or BUSY (i.e. halt the submission until
another command returns ... or the queue pressure builds if there are no
outstanding commands).

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agozfcp: add point-2-point support
<jejb@titanic.il.steeleye.com> [Mon, 11 Apr 2005 04:04:28 +0000 (23:04 -0500)]
zfcp: add point-2-point support

From: Andreas Herrmann <aherrman@de.ibm.com>

This patch mainly introduces support for point-2-point
topology.

From: Heiko Carstens <heiko.carstens@de.ibm.com>
From: Maxim Shchetynin <maxim@de.ibm.com>
From: Andreas Herrmann <aherrman@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] Convert i2o to compat_ioctl
<jejb@titanic.il.steeleye.com> [Mon, 11 Apr 2005 03:29:42 +0000 (22:29 -0500)]
[PATCH] Convert i2o to compat_ioctl

Also broken design in its compat handlers - CONFIG_COMPAT doesn't
mean that there should be no native ioctls.

Signed-off-by: Andi Kleen <ak@muc.de>
Reworked with comments from Markus Lidel by:
Christoph Hellwig <hch@infradead.org>

Rejections fixed and
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] kill old EH constants
<hch@lst.de> [Mon, 11 Apr 2005 03:19:25 +0000 (22:19 -0500)]
[PATCH] kill old EH constants

Fix up two drivers that incorrectly were using the old return values for
their new-style EH methods and kill off scsi_obsolete.h that defined the
constants.  The initio driver has all these constansts defined locally
and uses them internally, I'll fix that up some time later.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] scsi: remove meaningless scsi_cmnd->serial_number_at_timeout field
<htejun@gmail.com> [Sun, 3 Apr 2005 19:59:11 +0000 (14:59 -0500)]
[PATCH] scsi: remove meaningless scsi_cmnd->serial_number_at_timeout field

scsi_cmnd->serial_number_at_timeout doesn't serve any purpose
anymore.  All serial_number == serial_number_at_timeout tests
are always true in abort callbacks.  Kill the field.  Also, as
->pid always equals ->serial_number and ->serial_number
doesn't have any special meaning anymore, update comments
above ->serial_number accordingly.  Once we remove all uses of
this field from all lldd's, this field should go.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] scsi: remove unused scsi_cmnd->internal_timeout field
<htejun@gmail.com> [Sun, 3 Apr 2005 19:55:06 +0000 (14:55 -0500)]
[PATCH] scsi: remove unused scsi_cmnd->internal_timeout field

scsi_cmnd->internal_timeout field doesn't have any meaning
anymore.  Kill the field.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] remove outdated print_* functions
<hch@lst.de> [Sun, 3 Apr 2005 19:53:59 +0000 (14:53 -0500)]
[PATCH] remove outdated print_* functions

We have the scsi_print_* functions in the proper namespace for a long
time now and there weren't a lot users left.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] consolidate timeout defintions in scsi.h
<hch@lst.de> [Sun, 3 Apr 2005 19:52:44 +0000 (14:52 -0500)]
[PATCH] consolidate timeout defintions in scsi.h

Adapted from a patch in SuSE's kernel SRPM.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years agoupdates for CFQ oops fix
<jejb@titanic.il.steeleye.com> [Tue, 12 Apr 2005 21:38:09 +0000 (16:38 -0500)]
updates for CFQ oops fix

- add a comment to the device structure that the device_busy field
  is now protected by the request_queue->queue_lock
- null out sdev->request_queue after the queue is released to trap
  any (and there shouldn't be any) use after the queue is freed.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] fix NMI lockup with CFQ scheduler
<axboe@suse.de> [Tue, 12 Apr 2005 21:22:06 +0000 (16:22 -0500)]
[PATCH] fix NMI lockup with CFQ scheduler

The current problem seen is that the queue lock is actually in the
SCSI device structure, so when that structure is freed on device
release, we go boom if the queue tries to access the lock again.

The fix here is to move the lock from the scsi_device to the queue.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] finally fix 53c700 to use the generic iomem infrastructure
<jejb@mulgrave.(none)> [Sun, 3 Apr 2005 09:57:48 +0000 (03:57 -0600)]
[PATCH] finally fix 53c700 to use the generic iomem infrastructure

This driver has had it's own different infrastructure for doing this for
ages, but it's time it used the common one.

Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] zfcp: convert to compat_ioctl
<aherrman@de.ibm.com> [Sat, 2 Apr 2005 19:57:17 +0000 (13:57 -0600)]
[PATCH] zfcp: convert to compat_ioctl

Signed-off-by: Andreas Herrmann <aherrman@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] sg.c: update
<dougg@torque.net> [Sat, 2 Apr 2005 19:51:23 +0000 (13:51 -0600)]
[PATCH] sg.c: update

The attachment combines the most recent patch from
Yum Rayan <yum.rayan@gmail.com> (to reduce sg stack
usage), Adrian Bunk <bunk@stusta.de> (to fix check
after use) and me (fix elapsed time calculation
(duration) on ia64 machines).

I have modified the patch from Yum Rayan so kmalloc()
in sg_read() is only called for the (rare) code paths
that need them.

Changelog:
   - reduce stack usage in sg_ioctl() and sg_read()
   - fix check after use in sg_mmap()
   - hold duration internally in milliseconds and
     check current time later than held time

Signed-off-by: Douglas Gilbert <dougg@torque.net>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>
19 years ago[PATCH] md: remove a number of misleading calls to MD_BUG
NeilBrown [Sat, 16 Apr 2005 22:26:42 +0000 (15:26 -0700)]
[PATCH] md: remove a number of misleading calls to MD_BUG

The conditions that cause these calls to MD_BUG are not kernel bugs, just
oddities in what userspace is asking for.

Also convert analyze_sbs to return void, and the value it returned was
always 0.

Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] md: close a small race in md thread deregistration
NeilBrown [Sat, 16 Apr 2005 22:26:41 +0000 (15:26 -0700)]
[PATCH] md: close a small race in md thread deregistration

There is a tiny race when de-registering an MD thread, in that the thread
could disappear before it is set a SIGKILL, causing send_sig to have
problems.

This is most easily closed by holding tasklist_lock between enabling the
thread to exit (setting ->run to NULL) and telling it to exit.

(akpm: ick.  Needs to use kthread API and stop using signals)

Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] fbdev MAINTAINERS update
Benjamin Herrenschmidt [Sat, 16 Apr 2005 22:26:40 +0000 (15:26 -0700)]
[PATCH] fbdev MAINTAINERS update

This patch does the long overdue updates to MAINTAINERS file for aty128fb
and radeonfb.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] serial: fix comments in 8250.c
Russell King [Sat, 16 Apr 2005 22:26:39 +0000 (15:26 -0700)]
[PATCH] serial: fix comments in 8250.c

Fix the formatting of some comments in 8250.c, and add a note that the
register_serial / unregister_serial shouldn't be used in new code.

We do this here in preference to adding to linux/serial.h, since that is used
by a number of non-8250 drivers which pretend to be 8250.  It is not known
whether it would be appropriate to do so.

Signed-off-by: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] nfsd4: fix struct file leak
NeilBrown [Sat, 16 Apr 2005 22:26:38 +0000 (15:26 -0700)]
[PATCH] nfsd4: fix struct file leak

We were failing to close on an error path, resulting in a leak of struct files
which could take a v4 server down fairly quickly....  So call
nfs4_close_delegation instead of just open-coding parts of it.

Simplify the cleanup on delegation failure while we're at it.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] nfsd4: callback create rpc client returns
NeilBrown [Sat, 16 Apr 2005 22:26:38 +0000 (15:26 -0700)]
[PATCH] nfsd4: callback create rpc client returns

rpc_create_clnt and friends return errors, not NULL, on failure.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] nfsd: clear signals before exiting the nfsd() thread
NeilBrown [Sat, 16 Apr 2005 22:26:37 +0000 (15:26 -0700)]
[PATCH] nfsd: clear signals before exiting the nfsd() thread

Fixes the error "RPC: failed to contact portmap (errno -512)." when the server
later tries to unregister from the portmapper.

Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Signed-off-by: Neil Brown <neilb@cse.unsw.edu.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] jbd dirty buffer leak fix
akpm@osdl.org [Sat, 16 Apr 2005 22:26:36 +0000 (15:26 -0700)]
[PATCH] jbd dirty buffer leak fix

This fixes the lots-of-fsx-linux-instances-cause-a-slow-leak bug.

It's been there since 2.6.6, caused by:

ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mm4/broken-out/jbd-move-locked-buffers.patch

That patch moves under-writeout ordered-data buffers onto a separate journal
list during commit.  It took out the old code which was based on a single
list.

The old code (necessarily) had logic which would restart I/O against buffers
which had been redirtied while they were on the committing transaction's
t_sync_datalist list.  The new code only writes buffers once, ignoring
redirtyings by a later transaction, which is good.

But over on the truncate side of things, in journal_unmap_buffer(), we're
treating buffers on the t_locked_list as inviolable things which belong to the
committing transaction, and we just leave them alone during concurrent
truncate-vs-commit.

The net effect is that when truncate tries to invalidate a page whose buffers
are on t_locked_list and have been redirtied, journal_unmap_buffer() just
leaves those buffers alone.  truncate will remove the page from its mapping
and we end up with an anonymous clean page with dirty buffers, which is an
illegal state for a page.  The JBD commit will not clean those buffers as they
are removed from t_locked_list.  The VM (try_to_free_buffers) cannot reclaim
these pages.

The patch teaches journal_unmap_buffer() about buffers which are on the
committing transaction's t_locked_list.  These buffers have been written and
I/O has completed.  We can take them off the transaction and undirty them
within the context of journal_invalidatepage()->journal_unmap_buffer().

Acked-by: "Stephen C. Tweedie" <sct@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] drivers/infiniband/hw/mthca/mthca_main.c: remove an unused label
Roland Dreier [Sat, 16 Apr 2005 22:26:35 +0000 (15:26 -0700)]
[PATCH] drivers/infiniband/hw/mthca/mthca_main.c: remove an unused label

Correct unwinding in error path of mthca_init_icm().

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: add support for new MT25204 HCA
Roland Dreier [Sat, 16 Apr 2005 22:26:34 +0000 (15:26 -0700)]
[PATCH] IB/mthca: add support for new MT25204 HCA

Decouple table of HCA features from exact HCA device type.  Add a current FW
version field so we can warn when someone is using old FW.  Add support for
new MT25204 HCA.

Remove the warning about mem-free support, since it should be pretty solid at
this point.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: map context for RDMA responder in mem-free mode
Roland Dreier [Sat, 16 Apr 2005 22:26:34 +0000 (15:26 -0700)]
[PATCH] IB/mthca: map context for RDMA responder in mem-free mode

Fix RDMA in mem-free mode: we need to make sure that the RDMA context memory
is mapped for the HCA.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: update receive queue initialization for new HCAs
Roland Dreier [Sat, 16 Apr 2005 22:26:33 +0000 (15:26 -0700)]
[PATCH] IB/mthca: update receive queue initialization for new HCAs

Update initialization of receive queue to match new documentation.  This
change is required to support new MT25204 HCA.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: encapsulate mem-free check into mthca_is_memfree()
Roland Dreier [Sat, 16 Apr 2005 22:26:32 +0000 (15:26 -0700)]
[PATCH] IB/mthca: encapsulate mem-free check into mthca_is_memfree()

Clean up mem-free mode support by introducing mthca_is_memfree() function,
which encapsulates the logic of deciding if a device is mem-free.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: tweaks to mthca_cmd.c
Roland Dreier [Sat, 16 Apr 2005 22:26:31 +0000 (15:26 -0700)]
[PATCH] IB/mthca: tweaks to mthca_cmd.c

Minor tweaks to firmware command handling: kill off an unused get of a value,
and add a little more info to debug output.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: add fast memory region implementation
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:30 +0000 (15:26 -0700)]
[PATCH] IB/mthca: add fast memory region implementation

Implement fast memory regions (FMRs), where the driver writes directly into
the HCA's translation tables rather than requiring a firmware command.  For
Tavor, MTTs for FMR are separate from regular MTTs, and are reserved at driver
initialization.  This is done to limit the amount of virtual memory needed to
map the MTTs.  For Arbel, there's no such limitation, and all MTTs and MPTs
may be used for FMR or for regular MR.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: split MR key munging routines
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:30 +0000 (15:26 -0700)]
[PATCH] IB/mthca: split MR key munging routines

Split Tavor and Arbel/mem-free index<->hw key munging routines, so that FMR
implementation can call correct implementation without testing HCA type (which
it already knows).

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: add mthca_table_find() function
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:29 +0000 (15:26 -0700)]
[PATCH] IB/mthca: add mthca_table_find() function

Add mthca_table_find() function, which returns the lowmem address of an entry
in a mem-free HCA's context tables.  This will be used by the FMR
implementation.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: add SYNC_TPT firmware command
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:28 +0000 (15:26 -0700)]
[PATCH] IB/mthca: add SYNC_TPT firmware command

Add code for SYNC_TPT firmware command, which will be used by FMR
implementation.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: add mthca_write64_raw() for writing to MTT table directly
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:27 +0000 (15:26 -0700)]
[PATCH] IB/mthca: add mthca_write64_raw() for writing to MTT table directly

Add mthca_write64_raw() function, which will be used to write FMR entries that
are in ioremapped PCI memory.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: encapsulate MTT buddy allocator
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:26 +0000 (15:26 -0700)]
[PATCH] IB/mthca: encapsulate MTT buddy allocator

Encapsulate the buddy allocator used for MTT segments.  This cleans up the
code and also gets us ready to add FMR support.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: allow address handle creation in interrupt context
Roland Dreier [Sat, 16 Apr 2005 22:26:26 +0000 (15:26 -0700)]
[PATCH] IB/mthca: allow address handle creation in interrupt context

Make address handle verbs usable from interrupt context.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: fill in opcode field for send completions
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:25 +0000 (15:26 -0700)]
[PATCH] IB/mthca: fill in opcode field for send completions

Fill in missing fields in send completions.

Signed-off-by: Itamar Rabenstein <itamar@mellanox.co.il>
Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: fix MTT allocation in mem-free mode
Roland Dreier [Sat, 16 Apr 2005 22:26:24 +0000 (15:26 -0700)]
[PATCH] IB/mthca: fix MTT allocation in mem-free mode

Fix bug in MTT allocation in mem-free mode.

I misunderstood the MTT size value returned by the firmware -- it is really
the size of a single MTT entry, since mem-free mode does not segment the MTT
as the original firmware did.  This meant that our MTT addresses ended up
being off by a factor of 8.  This meant that our MTT allocations might
overlap, and so we could overwrite and corrupt earlier memory regions when
writing new MTT entries.

We fix this by always using our 64-byte MTT segment size.  This allows some
simplification of the code as well, since there's no reason to put the MTT
segment size in a variable -- we can always use our enum value directly.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: implement RDMA/atomic operations for mem-free mode
Roland Dreier [Sat, 16 Apr 2005 22:26:23 +0000 (15:26 -0700)]
[PATCH] IB/mthca: implement RDMA/atomic operations for mem-free mode

Add code to support RDMA and atomic send work requests in mem-free mode.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: fix format of CQ number for CQ events
Roland Dreier [Sat, 16 Apr 2005 22:26:22 +0000 (15:26 -0700)]
[PATCH] IB/mthca: fix format of CQ number for CQ events

CQ numbers are only 24 bits, so only print 6 hex digits and mask off reserved
part when reporting a CQ event.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: only free doorbell records in mem-free mode
Roland Dreier [Sat, 16 Apr 2005 22:26:21 +0000 (15:26 -0700)]
[PATCH] IB/mthca: only free doorbell records in mem-free mode

On error path, only free doorbell records if we're in mem-free mode.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: print assigned IRQ when interrupt test fails
Roland Dreier [Sat, 16 Apr 2005 22:26:21 +0000 (15:26 -0700)]
[PATCH] IB/mthca: print assigned IRQ when interrupt test fails

Print IRQ number when NOP command interrupt test fails to help debugging.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: release mutex on doorbell alloc error path
Roland Dreier [Sat, 16 Apr 2005 22:26:20 +0000 (15:26 -0700)]
[PATCH] IB/mthca: release mutex on doorbell alloc error path

Release mutex on error return path from mthca_alloc_db().

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: fix MR allocation error path
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:19 +0000 (15:26 -0700)]
[PATCH] IB/mthca: fix MR allocation error path

Fix error handling in MR allocation for mem-free mode: mthca_free must get an
MR index, not a key.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: allocate correct number of doorbell pages
Roland Dreier [Sat, 16 Apr 2005 22:26:18 +0000 (15:26 -0700)]
[PATCH] IB/mthca: allocate correct number of doorbell pages

Doorbell record pages are allocated in HCA page size chunks (always 4096
bytes), so we need to divide by 4096 and not PAGE_SIZE when figuring out how
many pages we'll need space for.

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: clean up mthca_dereg_mr()
Roland Dreier [Sat, 16 Apr 2005 22:26:17 +0000 (15:26 -0700)]
[PATCH] IB/mthca: clean up mthca_dereg_mr()

It's cleaner to kfree mthca_mr, and not rely on the fact that ib_mr is the
first field in mthca_mr.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: allow unaligned memory regions
Michael S. Tsirkin [Sat, 16 Apr 2005 22:26:16 +0000 (15:26 -0700)]
[PATCH] IB/mthca: allow unaligned memory regions

The first buffer of a memory region is not required to be page-aligned, so
don't return an error if it's not.

Signed-off-by: Michael S. Tsirkin <mst@mellanox.co.il>
Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
19 years ago[PATCH] IB/mthca: fix posting sends with immediate data
Roland Dreier [Sat, 16 Apr 2005 22:26:16 +0000 (15:26 -0700)]
[PATCH] IB/mthca: fix posting sends with immediate data

When posting a work request with immediate data, put the immediate data in the
immediate data field of the hardware's work request (rather than overwriting
the flags field).

Signed-off-by: Roland Dreier <roland@topspin.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>