]> git.hungrycats.org Git - linux/log
linux
22 years ago[PATCH] scsi/ppa.c cleanup and fixes (1/9)
Alexander Viro [Thu, 5 Feb 2004 01:44:13 +0000 (17:44 -0800)]
[PATCH] scsi/ppa.c cleanup and fixes (1/9)

* ppa.c fed through Lindent, functions unused elsewhere (by now -
  all of them) made static.

22 years ago[PATCH] scsi/imm.c cleanup and fixes (8/8)
Alexander Viro [Thu, 5 Feb 2004 01:43:11 +0000 (17:43 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (8/8)

* switched to proper parport_driver.

22 years ago[PATCH] scsi/imm.c cleanup and fixes (7/8)
Alexander Viro [Thu, 5 Feb 2004 01:41:59 +0000 (17:41 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (7/8)

* switched to use ->hostdata to get imm_struct from Scsi_Host, end
of messing with imm_hosts[]

22 years ago[PATCH] scsi/imm.c cleanup and fixes (6/8)
Alexander Viro [Thu, 5 Feb 2004 01:40:55 +0000 (17:40 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (6/8)

* fixed missing wakeups in imm_pb_claim()/imm_wakeup() - if the
former had been called just as current holder of port was giving it up,
we could set "I'm waiting" flag too late.  Cleaned up the timeout logics.

22 years ago[PATCH] scsi/imm.c cleanup and fixes (5/8)
Alexander Viro [Thu, 5 Feb 2004 01:39:40 +0000 (17:39 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (5/8)

* starting to kill imm_detect - we move the "probe a single port"
logics into a separate function and shift scanning into imm_driver_init().
Later that will give us a parport_driver ->attach().

22 years ago[PATCH] scsi/imm.c cleanup and fixes (4/8)
Alexander Viro [Thu, 5 Feb 2004 01:38:20 +0000 (17:38 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (4/8)

* killed include of scsi_module.h and switched from scsi_register()
to scsi_alloc_host().  We still keep the old detect logics at that point,
it will be gone later.

22 years ago[PATCH] scsi/imm.c cleanup and fixes (3/8)
Alexander Viro [Thu, 5 Feb 2004 01:36:55 +0000 (17:36 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (3/8)

* killed imm_pb_claim() call in imm_init()

22 years ago[PATCH] scsi/imm.c cleanup and fixes (2/8)
Alexander Viro [Thu, 5 Feb 2004 01:35:29 +0000 (17:35 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (2/8)

* switched imm to passing around references to imm_struct instead
of host numbers.

22 years ago[PATCH] scsi/imm.c cleanup and fixes (1/8)
Alexander Viro [Thu, 5 Feb 2004 01:34:01 +0000 (17:34 -0800)]
[PATCH] scsi/imm.c cleanup and fixes (1/8)

* imm.c fed through Lindent, functions unused elsewhere (by now -
  all of them) made static.

22 years ago[PATCH] paride cleanup and fixes (25/25)
Alexander Viro [Thu, 5 Feb 2004 01:32:24 +0000 (17:32 -0800)]
[PATCH] paride cleanup and fixes (25/25)

paride has a ->private field for use by low-level drivers.  It's not
64bit-clean - it's int, but is used to store a pointer (bpck6).
Fix: replace with unsigned long, update typecasts in users.

22 years ago[PATCH] paride cleanup and fixes (24/25)
Alexander Viro [Thu, 5 Feb 2004 01:30:46 +0000 (17:30 -0800)]
[PATCH] paride cleanup and fixes (24/25)

pd_identify() converted to pd_special_command().

22 years ago[PATCH] paride cleanup and fixes (23/25)
Alexander Viro [Thu, 5 Feb 2004 01:28:59 +0000 (17:28 -0800)]
[PATCH] paride cleanup and fixes (23/25)

pd_init_units() merged into pd_detect(), pd_drive_count made local to
pd_detect().

22 years ago[PATCH] paride cleanup and fixes (22/25)
Alexander Viro [Thu, 5 Feb 2004 01:27:09 +0000 (17:27 -0800)]
[PATCH] paride cleanup and fixes (22/25)

Now disk->present is not needed - we can always check for disk->gd.
Field removed, pd_probe_drive() became void(struct pd_unit *), pd_detect()
cleaned up,

22 years ago[PATCH] paride cleanup and fixes (21/25)
Alexander Viro [Thu, 5 Feb 2004 01:25:22 +0000 (17:25 -0800)]
[PATCH] paride cleanup and fixes (21/25)

Moved allocation and setup (but not adding) of gendisk into
pd_probe_drive().

22 years ago[PATCH] paride cleanup and fixes (20/25)
Alexander Viro [Thu, 5 Feb 2004 01:23:39 +0000 (17:23 -0800)]
[PATCH] paride cleanup and fixes (20/25)

Switched everything except pd_identify() to use of special requests.
New helper: pd_special_command() - builds special request, feeds it
into queue and waits for completion.
Moved pi_connect()/pi_disconnect() to the beginning/end of pi_identify.

22 years ago[PATCH] paride cleanup and fixes (19/25)
Alexander Viro [Thu, 5 Feb 2004 01:22:02 +0000 (17:22 -0800)]
[PATCH] paride cleanup and fixes (19/25)

Reordered stuff
made pd_ready() inlined
renamed ps_timeout to pd_timeout
renamed ps_tq to fsm_tq
renamed ps_set_intr to fsm_schedule
removed unused pd_wait_open.

22 years ago[PATCH] paride cleanup and fixes (18/25)
Alexander Viro [Thu, 5 Feb 2004 01:20:20 +0000 (17:20 -0800)]
[PATCH] paride cleanup and fixes (18/25)

Moved calls of ->connect and ->disconnect into run_fsm().  The only
change is that now do_pd_start_io() is ran when we are connected.

22 years ago[PATCH] paride cleanup and fixes (17/25)
Alexander Viro [Thu, 5 Feb 2004 01:18:42 +0000 (17:18 -0800)]
[PATCH] paride cleanup and fixes (17/25)

Equivalent transformation:
now that all phase functions are called claimed, we can
simplify the logics around claiming - new variable (pd_claimed)
keeps track of claim/unclaim we'd done.  We check it in the
beginning of the loop and do claiming if it's not set.

22 years ago[PATCH] paride cleanup and fixes (16/25)
Alexander Viro [Thu, 5 Feb 2004 01:16:58 +0000 (17:16 -0800)]
[PATCH] paride cleanup and fixes (16/25)

Equivalent transformation:
* phase = do_pd_io replaced with phase = NULL, corresponding
  check + open-coded equivalent added in run_fsm loop.
* do_pd_io() is gone.

22 years ago[PATCH] paride cleanup and fixes (15/25)
Alexander Viro [Thu, 5 Feb 2004 01:15:22 +0000 (17:15 -0800)]
[PATCH] paride cleanup and fixes (15/25)

Equivalent transformation:
next_request() call expanded.

22 years ago[PATCH] paride cleanup and fixes (14/25)
Alexander Viro [Thu, 5 Feb 2004 01:14:00 +0000 (17:14 -0800)]
[PATCH] paride cleanup and fixes (14/25)

Equivalent transformation:
call of pi_do_claimed() in run_fsm() expanded; recursive
call of run_fsm() replaced with loop.

22 years ago[PATCH] paride cleanup and fixes (13/25)
Alexander Viro [Thu, 5 Feb 2004 01:12:32 +0000 (17:12 -0800)]
[PATCH] paride cleanup and fixes (13/25)

Equivalent transformation:
pi_unclaim() moved from phase functions to run_fsm().

22 years ago[PATCH] paride cleanup and fixes (12/25)
Alexander Viro [Thu, 5 Feb 2004 01:11:09 +0000 (17:11 -0800)]
[PATCH] paride cleanup and fixes (12/25)

Equivalent transformation:
* calls of pi_disconnect() in phase functions expanded.
* since both do_pd_read_start() and do_pd_write_start()
  are always called claimed (as the matter of fact,
  all phase functions except do_pd_io() are), we can
  replace call of pi_connect() with direct method call -
  pi_claim() is no-op there.

22 years ago[PATCH] paride cleanup and fixes (11/25)
Alexander Viro [Thu, 5 Feb 2004 01:09:53 +0000 (17:09 -0800)]
[PATCH] paride cleanup and fixes (11/25)

Equivalent transformation:
* new type - enum action.  phase functions return it now.
  run_fsm() looks at the result of phase() and does the
  corresponding action - next_request(), pi_do_claimed()
  or ps_set_intr().

22 years ago[PATCH] paride cleanup and fixes (10/25)
Alexander Viro [Thu, 5 Feb 2004 01:08:47 +0000 (17:08 -0800)]
[PATCH] paride cleanup and fixes (10/25)

Equivalent transformation:
* ps_continuation renamed - it's "phase" now.
* run_fsm() does what ps_tq_int() used to do
* calls of pi_do_claimed(..., func) replaced with
  phase = func; pi_do_claimed(..., run_fsm);
Now all phase functions are called from run_fsm() and we
can start pulling the common code into it.

22 years ago[PATCH] paride cleanup and fixes (9/25)
Alexander Viro [Thu, 5 Feb 2004 01:07:39 +0000 (17:07 -0800)]
[PATCH] paride cleanup and fixes (9/25)

Equivalent transformation:
do_pd_reqeust1() calls lambda-expanded, function killed.
resulting call of ps_set_intr() in next_request() has been
moved past dropping pd_lock.

22 years ago[PATCH] paride cleanup and fixes (8/25)
Alexander Viro [Thu, 5 Feb 2004 01:06:25 +0000 (17:06 -0800)]
[PATCH] paride cleanup and fixes (8/25)

Equivalent transformation:
instead of having do_pd_request1() calling do_pd_io() via
pi_do_claimed() and do_pd_io() calling do_pd_io_start() via
ps_set_intr(), do it the other way round.

22 years ago[PATCH] paride cleanup and fixes (7/25)
Alexander Viro [Thu, 5 Feb 2004 01:05:25 +0000 (17:05 -0800)]
[PATCH] paride cleanup and fixes (7/25)

Equivalent transformation:
Handling of bogus requests taken into do_pd_io_start() - we
call next_request(0) as we would for any other error.  We
need to call pi_unclaim() before that - on other paths that
would happen as a side effect of pi_disconnect().
Setting variables moved into do_pd_io_start() - they don't
need pd_lock being held, so we can shift that outside.

22 years ago[PATCH] paride cleanup and fixes (6/25)
Alexander Viro [Thu, 5 Feb 2004 01:04:29 +0000 (17:04 -0800)]
[PATCH] paride cleanup and fixes (6/25)

Equivalent transformation:
* new function: do_pd_io_start().  Checks pd_cmd and calls
  do_pd_read_start() or do_pd_write_start().
* do_pd_read() and do_pd_write() merged into do_pd_io(), which
  schedules execution of do_pd_io_start().

22 years ago[PATCH] paride cleanup and fixes (5/25)
Alexander Viro [Thu, 5 Feb 2004 01:03:38 +0000 (17:03 -0800)]
[PATCH] paride cleanup and fixes (5/25)

pd_busy gone - we can check for pd_req != NULL instead.

22 years ago[PATCH] paride cleanup and fixes (4/25)
Alexander Viro [Thu, 5 Feb 2004 01:02:45 +0000 (17:02 -0800)]
[PATCH] paride cleanup and fixes (4/25)

Logics around "if not ready, schedule the same step again" taken
out of ps_tq_intr() and into the steps that need it.

22 years ago[PATCH] paride cleanup and fixes (3/25)
Alexander Viro [Thu, 5 Feb 2004 01:02:01 +0000 (17:02 -0800)]
[PATCH] paride cleanup and fixes (3/25)

Removed junk (see the analysis in the previous patch), removed
ps_set_intr() arguments.

22 years ago[PATCH] paride cleanup and fixes (2/25)
Alexander Viro [Thu, 5 Feb 2004 01:01:27 +0000 (17:01 -0800)]
[PATCH] paride cleanup and fixes (2/25)

Preparation to crapectomy - expanded include of pseudo.h, split
the part of pd_do_request() into pd_do_request1(), leaving in pd_do_request()
only the check that FSM is not running and call of pd_do_request1().
Added locking analysis - it's in drivers/block/paride/Transition-notes
and it will be used on the next step.  Even though the subsequent steps are
small, they rely on properties of the damn thing and proof of these properties
is hairy.  Amazing perversions people manage to produce when writing a trivial
FSM - the code is very convoluted for no good reason...

22 years ago[PATCH] paride cleanup and fixes (1/25)
Alexander Viro [Thu, 5 Feb 2004 01:00:52 +0000 (17:00 -0800)]
[PATCH] paride cleanup and fixes (1/25)

* new helper: pi_schedule_claimed() - pi_do_claimed() sans the call
  of continuation if we don't have to wait; returns 1 if we don't
  have to wait and 0 otherwise.

22 years ago[PATCH] parport fixes (6/6)
Alexander Viro [Thu, 5 Feb 2004 01:00:21 +0000 (17:00 -0800)]
[PATCH] parport fixes (6/6)

daisy.c used to access the topology list with no locking whatsoever.
Protected by spinlock, fixed numerous bugs in handling of the single-linked
list (I'm not kidding - just take a look at the old parport_daisy_fini()
and weep; it is supposed to go through a simple list and remove all entries
that satisfy a condition).

22 years ago[PATCH] parport fixes (5/6)
Alexander Viro [Thu, 5 Feb 2004 00:59:52 +0000 (16:59 -0800)]
[PATCH] parport fixes (5/6)

bw-qcam.c turned into proper parport driver.  Instead of (racy)
scanning the list of ports we use ->attach() and ->detach().

22 years ago[PATCH] parport fixes (4/6)
Alexander Viro [Thu, 5 Feb 2004 00:59:29 +0000 (16:59 -0800)]
[PATCH] parport fixes (4/6)

parport_gsc.c turned into proper parisc driver; instead of scanning
the list of ports upon rmmod in search of ones that had been created by us,
we do cleanup where it belongs - in parisc driver ->remove().

22 years ago[PATCH] parport fixes (3/6)
Alexander Viro [Thu, 5 Feb 2004 00:59:07 +0000 (16:59 -0800)]
[PATCH] parport fixes (3/6)

A bunch of parport_enumerate() users was duplicating parport_find_...()
without proper locking.  Replaced with use of appropriate helpers, races closed.

22 years ago[PATCH] parport fixes (2/6)
Alexander Viro [Thu, 5 Feb 2004 00:58:49 +0000 (16:58 -0800)]
[PATCH] parport fixes (2/6)

We use a new mutex to protect all additions/removals of drivers and
ports.  That cures a lot of insanity:
* driver removals can't hit us in the middle of attach_driver_chain().
Old code simply dies on that.
* port removals can't hit us in the middle of driver registration.
Again, old code dies on that.
* driver ->detach() is allowed to block now.
* we are guaranteed that by the time when parport_unregister_driver()
returns, all ->detach() calls are finished.  Old code did _not_ guarantee
that (read: was inherently racy since rmmod of driver could race with port
removal and get driver->detach(port) called after the module was gone).
* we are guaranteed that driver->attach(port) won't be called
more than once.  With the old code that was a matter of luck.
* removed piles and piles of braindead code.

22 years ago[PATCH] parport fixes (1/6)
Alexander Viro [Thu, 5 Feb 2004 00:58:34 +0000 (16:58 -0800)]
[PATCH] parport fixes (1/6)

Current tree has all allocated ports on portlist.  However,
most of the portlist users assume that we only have announced ports
there and break badly if they happen to see the list after port driver
has created a port (parport_register_port()) but before it finishes
the setup (and calls parport_announce_port()).
The only place that wants to see all allocated port is
parport_register_port() itself and only to pick the first unused
port number.
We add a new list (all_ports) and put ports there when allocated;
that list is kept ordered by port->number.  Ports are placed on portlist
only by parport_announce_port().
Gobs of shite in parport_register_port() removed, pile of races
closed...

22 years agoMerge bk://gkernel.bkbits.net/libata-2.5
Linus Torvalds [Thu, 5 Feb 2004 00:55:02 +0000 (16:55 -0800)]
Merge bk://gkernel.bkbits.net/libata-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years agoMerge redhat.com:/spare/repo/linux-2.5
Jeff Garzik [Thu, 5 Feb 2004 01:03:42 +0000 (20:03 -0500)]
Merge redhat.com:/spare/repo/linux-2.5
into redhat.com:/spare/repo/libata-2.5

22 years agoMerge http://typhoon.bkbits.net/typhoon-2.5
Jeff Garzik [Thu, 5 Feb 2004 00:19:14 +0000 (19:19 -0500)]
Merge http://typhoon.bkbits.net/typhoon-2.5
into redhat.com:/spare/repo/net-drivers-2.5

22 years ago[wireless atmel] update
Simon Kelley [Thu, 5 Feb 2004 00:04:45 +0000 (19:04 -0500)]
[wireless atmel] update

Add PCI device support - there are atmel_pci.ko and atmel_cs.ko
modules and a library module called atmel.ko

Tweak the PCMCIA card -> firmware table for new cards.

Fix workarounds for uniquely broken 3com cards, which were
rendered unuable by the 0.9 changes.

Bump version to 0.91

22 years ago[PATCH] Handle an old acenic card
Matthew Wilcox [Wed, 4 Feb 2004 23:54:45 +0000 (18:54 -0500)]
[PATCH] Handle an old acenic card

Hi Jes.  I really do have a Rev. 5 card, and it works much better treated
as a Tigon I than as a Tigon II.

22 years ago[wireless airo] Add support for mini-pci based cards
Javier Achirica [Wed, 4 Feb 2004 23:54:37 +0000 (18:54 -0500)]
[wireless airo] Add support for mini-pci based cards

22 years ago[PATCH] hamradio driver fixes
Jeroen Vreeken [Wed, 4 Feb 2004 23:48:25 +0000 (18:48 -0500)]
[PATCH] hamradio driver fixes

I send these a while ago but didn't get a reply.
Can you apply these patches for the hamradio drivers?

details:

+ scc.c usage of rtnl lock and register_netdev.
+ scc.c properly initialize timers.
+ remove cli() call in 6pack.c
+ fix init of reused devices and remove cli()
+ scc.c statistics

22 years ago[PATCH] bmac network driver update
Benjamin Herrenschmidt [Wed, 4 Feb 2004 23:48:16 +0000 (18:48 -0500)]
[PATCH] bmac network driver update

Hi !

I didn't include that with the bitkeeper based PowerMac updates so Jeff
can assume his role as maintainer/filter of network driver updates :)

Here's my latest update to this pmac-only driver, please apply if you
are ok with it. The driver now uses the macio infrastructure for
detection & power management notifications, getting into sysfs at
the same time.

Ben.

# This is a BitKeeper generated diff -Nru style patch.
#
# ChangeSet
#   2004/01/23 09:38:35+11:00 benh@kernel.crashing.org
#   Update PowerMac "bmac" driver to the "macio" device so it fits in sysfs
#   and gets proper power management ordering
#
# drivers/net/bmac.c
#   2004/01/23 09:38:22+11:00 benh@kernel.crashing.org +185 -203
#   Update PowerMac "bmac" driver to the "macio" device so it fits in sysfs
#   and gets proper power management ordering
#

22 years ago[PATCH] [2.6] Update forcedeth to 0.23
Carl-Daniel Hailfinger [Wed, 4 Feb 2004 23:48:08 +0000 (18:48 -0500)]
[PATCH] [2.6] Update forcedeth to 0.23

This is a multi-part message in MIME format.

22 years agoMerge redhat.com:/spare/repo/netdev-2.6/tulip
Jeff Garzik [Wed, 4 Feb 2004 15:43:32 +0000 (10:43 -0500)]
Merge redhat.com:/spare/repo/netdev-2.6/tulip
into redhat.com:/spare/repo/net-drivers-2.5

22 years agoMerge redhat.com:/spare/repo/netdev-2.6/8139too
Jeff Garzik [Wed, 4 Feb 2004 15:41:32 +0000 (10:41 -0500)]
Merge redhat.com:/spare/repo/netdev-2.6/8139too
into redhat.com:/spare/repo/net-drivers-2.5

22 years ago[PATCH] use cramfs as an initrd
James Bottomley [Wed, 4 Feb 2004 05:51:43 +0000 (21:51 -0800)]
[PATCH] use cramfs as an initrd

Now that Al Viro fixed cramfs, it works beautifully as an initrd
filesystem.

So finally plumb it in.

22 years agoMerge davem@nuts.davemloft.net:/disk1/BK/sparc-2.6
David S. Miller [Wed, 4 Feb 2004 04:40:58 +0000 (20:40 -0800)]
Merge davem@nuts.davemloft.net:/disk1/BK/sparc-2.6
into kernel.bkbits.net:/home/davem/sparc-2.6

22 years ago[SPARC64}: Fix ultra-III and later support of new-style SILO booting.
David S. Miller [Wed, 4 Feb 2004 03:33:43 +0000 (19:33 -0800)]
[SPARC64}: Fix ultra-III and later support of new-style SILO booting.

22 years ago[PATCH] H8/300 support update (3/3): bitops
Yoshinori Sato [Wed, 4 Feb 2004 03:24:21 +0000 (19:24 -0800)]
[PATCH] H8/300 support update (3/3): bitops

o Cleanup reduced and faster code

22 years ago[PATCH] H8/300 support update (2/3): compiler warnings
Yoshinori Sato [Wed, 4 Feb 2004 03:24:09 +0000 (19:24 -0800)]
[PATCH] H8/300 support update (2/3): compiler warnings

o gcc-3.4 warning fix.

22 years ago[PATCH] H8/300 support update (1/3): obsolete header
Yoshinori Sato [Wed, 4 Feb 2004 03:23:57 +0000 (19:23 -0800)]
[PATCH] H8/300 support update (1/3): obsolete header

o Deleted obsolute header include

22 years agoMerge bk://linux-scsi.bkbits.net/scsi-for-linus-2.6
Linus Torvalds [Wed, 4 Feb 2004 03:20:10 +0000 (19:20 -0800)]
Merge bk://linux-scsi.bkbits.net/scsi-for-linus-2.6
into home.osdl.org:/home/torvalds/v2.5/linux

22 years agoMerge bk://gkernel.bkbits.net/net-drivers-2.5
Linus Torvalds [Wed, 4 Feb 2004 03:16:43 +0000 (19:16 -0800)]
Merge bk://gkernel.bkbits.net/net-drivers-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[PATCH] Fix block device inode list corruptions
Stephen C. Tweedie [Wed, 4 Feb 2004 02:57:45 +0000 (18:57 -0800)]
[PATCH] Fix block device inode list corruptions

I've been chasing a weird SELinux bug which shows up mostly when doing
installs of a dev-* rpm (ie. creating and overwriting lots of block
device inodes), but which I've also seen when doing mkinitrd.

It turned out not to be an SELinux problem at all, but a core VFS
S_ISBLK bug.  It seems that SELinux simply widens the race window.

The code at fault is fs/fs-writeback.c:__mark_inode_dirty():

/*
 * Only add valid (hashed) inodes to the superblock's
 * dirty list.  Add blockdev inodes as well.
 */
if (!S_ISBLK(inode->i_mode)) {
if (hlist_unhashed(&inode->i_hash))
goto out;
if (inode->i_state & (I_FREEING|I_CLEAR))
goto out;
}

The "I_FREEING|I_CLEAR" condition was added after the ISBLK/unhashed
tests were already in the source, but I can't see any reason why we'd
want the I_FREEING test not to apply to block devices.  And indeed, this
results in all sorts of inode list corruptions.  Simply moving the
I_FREEING|I_CLEAR test out of the protection of the S_ISBLK() condition
fixes things entirely.

The existing 2.6 kernel will reliably fail on me in about 2 seconds once
"rpm -Uvh --force dev*.rpm" starts its actual installation of the new
inodes.  With the patch below I can't reproduce it at all.

22 years ago[PATCH] Fix ptrace in the vsyscall dso area
Andrew Morton [Wed, 4 Feb 2004 02:54:15 +0000 (18:54 -0800)]
[PATCH] Fix ptrace in the vsyscall dso area

From: Roland McGrath <roland@redhat.com>

The #include is the part of this patch that matters, so the #ifdef below
works.

The rest of the patch removes gratuitous duplication due to some strange
aversion to concision in the presence of #ifdef, the kind that is all too
common, utterly pointless, and error prone.

22 years ago[PATCH] Altix update: pci_bus_cvlink.c fixes
Andrew Morton [Wed, 4 Feb 2004 02:54:06 +0000 (18:54 -0800)]
[PATCH] Altix update: pci_bus_cvlink.c fixes

From: Pat Gefre <pfg@sgi.com>

rch/ia64/sn/io/machvec/pci_bus_cvlink.c
    Some pretty-print mods
    Use pin instead of slot for dma_flush init

22 years ago[PATCH] Altix update: pci_bus_cvlink.c fixes
Andrew Morton [Wed, 4 Feb 2004 02:53:57 +0000 (18:53 -0800)]
[PATCH] Altix update: pci_bus_cvlink.c fixes

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/machvec/pci_bus_cvlink.c
    Couple of checks for kmalloc <= 0 were fixed
    Some __init and static fixes

22 years ago[PATCH] Altix update: irq fixes
Andrew Morton [Wed, 4 Feb 2004 02:53:48 +0000 (18:53 -0800)]
[PATCH] Altix update: irq fixes

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/kernel/irq.c
    Need to get the cpu from the passed in pcibr struct
    Made the interrupt list static and gave it a better name - credit jes
    Some lindent'isms
    Took out some code that isn't used ..... yet

22 years ago[PATCH] Altix update: remove pcibr_intr_func()
Andrew Morton [Wed, 4 Feb 2004 02:53:39 +0000 (18:53 -0800)]
[PATCH] Altix update: remove pcibr_intr_func()

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/sn2/pcibr/pcibr_intr.c
    Kill pcibr_intr_func()

22 years ago[PATCH] Altix update: VGA, keyboard, other changes
Andrew Morton [Wed, 4 Feb 2004 02:53:30 +0000 (18:53 -0800)]
[PATCH] Altix update: VGA, keyboard, other changes

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/kernel/setup.c

    If generic enabled legacy VGA or kbd - disable them
    Slightly different check for work arounds and only do it once
    If there is no klconfig info and we are in the simulator - ignore it
    Update the pxm_to_nasid() routine. It failed for SP configurations and some
    SMP configurations where M-bricks used pxm numbers lower that the first
    c-brick.
    If we don't find the cpu for pxm, search the memblks for it.

22 years ago[PATCH] Altix update: early_probe_for_widget() improvement
Andrew Morton [Wed, 4 Feb 2004 02:53:22 +0000 (18:53 -0800)]
[PATCH] Altix update: early_probe_for_widget() improvement

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/sn2/ml_iograph.c
    Better code for early_probe_for_widget()

22 years ago[PATCH] Altix update: pcibr_invalidate_ate check
Andrew Morton [Wed, 4 Feb 2004 02:53:13 +0000 (18:53 -0800)]
[PATCH] Altix update: pcibr_invalidate_ate check

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/sn2/pcibr/pcibr_ate.c
    check for pcibr_invalidate_ate 0 used for debugging

22 years ago[PATCH] Altix update: io changes
Andrew Morton [Wed, 4 Feb 2004 02:53:04 +0000 (18:53 -0800)]
[PATCH] Altix update: io changes

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/platform_init/sgi_io_init.c
   use numionodes instead of numnodes

arch/ia64/sn/io/sn2/klconflib.c
    find_lboard changes - generalized a number of the interface funcs

arch/ia64/sn/io/sn2/klgraph.c
    call the more general find_lboard funcs

arch/ia64/sn/io/sn2/ml_iograph.c
    call the more general lboard funcs

arch/ia64/sn/io/sn2/module.c
    lboard changes
    mod for headless/memless nodes

arch/ia64/sn/io/sn2/pcibr/pcibr_dvr.c
    isIO9 mod

arch/ia64/sn/kernel/setup.c
    headless/memless mod

include/asm-ia64/sn/klconfig.h
    generalized find_lboard funs

22 years ago[PATCH] Altix update: add MINIMAL_ATE_FLAG
Andrew Morton [Wed, 4 Feb 2004 02:52:55 +0000 (18:52 -0800)]
[PATCH] Altix update: add MINIMAL_ATE_FLAG

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/machvec/pci_dma.c
    Add MINIMAL_ATE_FLAGS() macro usage

include/asm-ia64/sn/pci/pcibr.h
    Add MINIMAL_ATE_FLAG() macro

22 years ago[PATCH] Altix update: misc changes
Andrew Morton [Wed, 4 Feb 2004 02:52:46 +0000 (18:52 -0800)]
[PATCH] Altix update: misc changes

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/sn2/pcibr/pcibr_dvr.c
    extern for pcibr_rrb_alloc_more()

include/asm-ia64/sn/pci/pcibr_private.h
    more unsigned to unsigned int

22 years ago[PATCH] Altix update: small cleanups
Andrew Morton [Wed, 4 Feb 2004 02:52:37 +0000 (18:52 -0800)]
[PATCH] Altix update: small cleanups

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/io.c@1.12
    misc code cleanup
    no parens on returns
    remove __psunsigned_t

arch/ia64/sn/io/sn2/ml_iograph.c@1.18
    no parens on returns

arch/ia64/sn/io/sn2/pcibr/pcibr_dvr.c@1.25
    no parens on returns
    use -errno

arch/ia64/sn/io/sn2/pcibr/pcibr_hints.c@1.8
    no parens on returns

arch/ia64/sn/io/sn2/pcibr/pcibr_intr.c@1.15
    no parens on returns
    unsigned to unsigned int

arch/ia64/sn/io/sn2/pcibr/pcibr_rrb.c@1.13
    no parens on returns
    use -errno

arch/ia64/sn/io/sn2/pcibr/pcibr_slot.c@1.18
    no parens on returns
    use -errno

arch/ia64/sn/io/sn2/pciio.c@1.14
    oom handling
    removed ASSERT and funcs not needed

arch/ia64/sn/io/sn2/pic.c@1.13
    use -errno

arch/ia64/sn/io/sn2/shuberror.c@1.12
    add delay.h

arch/ia64/sn/io/sn2/xbow.c@1.10
    include file clean up

arch/ia64/sn/io/sn2/xtalk.c@1.9
    misc code cleanup
    no parens on returns

arch/ia64/sn/io/xswitch.c@1.11
    misc code cleanup

arch/ia64/sn/kernel/bte.c@1.5
    include file clean up

arch/ia64/sn/kernel/mca.c@1.8
    include file clean up

arch/ia64/sn/kernel/probe.c@1.5
    include file clean up

arch/ia64/sn/kernel/sn2/prominfo_proc.c@1.3
    include file clean up

arch/ia64/sn/kernel/sn2/sn2_smp.c@1.9
    include file clean up

arch/ia64/sn/kernel/sn2/sn_proc_fs.c@1.4
    include file clean up

drivers/char/sn_serial.c@1.3
    include file clean up

include/asm-ia64/sn/addrs.h@1.8
    include file clean up
    remove __psunsigned_t

include/asm-ia64/sn/alenlist.h@1.8
    unsigned to unsigned int

include/asm-ia64/sn/arch.h@1.7
    include file clean up

include/asm-ia64/sn/bte.h@1.6
    include file clean up

include/asm-ia64/sn/clksupport.h@1.8
    include file clean up

include/asm-ia64/sn/driver.h@1.5
    remove __psunsigned_t

include/asm-ia64/sn/hcl.h@1.13
    include file clean up

include/asm-ia64/sn/hcl_util.h@1.7
    include file clean up

include/asm-ia64/sn/hwgfs.h@1.5
    include file clean up

include/asm-ia64/sn/iograph.h@1.9
    include file clean up

include/asm-ia64/sn/klconfig.h@1.14
    remove __psunsigned_t

include/asm-ia64/sn/kldir.h@1.5
    remove __psunsigned_t
    include file clean up

include/asm-ia64/sn/module.h@1.12
    include file clean up

include/asm-ia64/sn/nodepda.h@1.14
    include file clean up

include/asm-ia64/sn/pci/bridge.h@1.13
    uchar_t to unsigned char

include/asm-ia64/sn/pci/pcibr_private.h@1.20
    move PVnnnnnn's

include/asm-ia64/sn/pci/pciio.h@1.14
    unsigned to unsigned int

include/asm-ia64/sn/pci/pciio_private.h@1.10
    unsigned to unsigned int

include/asm-ia64/sn/pda.h@1.9
    include file clean up

include/asm-ia64/sn/pio.h@1.7
    include file clean up
    ulong to unsigned long

include/asm-ia64/sn/sgi.h@1.11
    include file clean up
    move PVnnnnnn's to here

include/asm-ia64/sn/sn2/arch.h@1.6
    include file clean up

include/asm-ia64/sn/sn2/sn_private.h@1.12
    include file clean up
    remove __psunsigned_t

include/asm-ia64/sn/sn_cpuid.h@1.8
    include file clean up

include/asm-ia64/sn/sn_private.h@1.6
    include file clean up

include/asm-ia64/sn/types.h@1.6
    include file clean up
    remove __psunsigned_t

include/asm-ia64/sn/vector.h@1.6
    include file clean up

include/asm-ia64/sn/xtalk/xbow.h@1.9
    include file clean up
    misc code cleanup

include/asm-ia64/sn/xtalk/xtalk.h@1.12
    unsigned to unsigned int

include/asm-ia64/sn/xtalk/xwidget.h@1.8
    unsigned to unsigned int

22 years ago[PATCH] Altix update: various, mainly cleanups
Andrew Morton [Wed, 4 Feb 2004 02:52:26 +0000 (18:52 -0800)]
[PATCH] Altix update: various, mainly cleanups

From: Pat Gefre <pfg@sgi.com>

arch/ia64/sn/io/machvec/pci_bus_cvlink.c
    Changes for new pcireg_ interfaces
    pcibr reorg
    Some code cleanup/reorg

arch/ia64/sn/io/machvec/pci_dma.c
    IS_PCIA64() not needed

arch/ia64/sn/io/sn2/ml_iograph.c
    new pcireg_ interface

arch/ia64/sn/io/sn2/pcibr/pcibr_ate.c
    code reorg/clean up

arch/ia64/sn/io/sn2/pcibr/pcibr_config.c
    code reorg/cleanup

arch/ia64/sn/io/sn2/pcibr/pcibr_dvr.c
    reorg/cleanup

arch/ia64/sn/io/sn2/pcibr/pcibr_error.c
    reorg/cleanup

arch/ia64/sn/io/sn2/pcibr/pcibr_intr.c
    reorg/cleanup

arch/ia64/sn/io/sn2/pcibr/pcibr_reg.c
    Fixed the interface to these functions - one call/data type

arch/ia64/sn/io/sn2/pcibr/pcibr_rrb.c
    reorg/cleanup

arch/ia64/sn/io/sn2/pcibr/pcibr_slot.c
    reorg/cleanup

arch/ia64/sn/io/sn2/pciio.c
    removed unused functions

arch/ia64/sn/io/sn2/pic.c
    reorg/cleanup

arch/ia64/sn/kernel/irq.c
    IS_PIC_SOFT not needed
    mod for new pcireg_ interfaces

include/asm-ia64/sn/module.h
    nodes/geoid[] -> MAX_SLABS

include/asm-ia64/sn/pci/bridge.h
    IS_[X]BRIDGE not needed

include/asm-ia64/sn/pci/pci_bus_cvlink.h
    SET_PCIA64 and IS_PCIA64 not needed
    isa64, dma_buf_sync, xbow_buf_sync gone

include/asm-ia64/sn/pci/pcibr.h
    mostly cleanup
    some reorg mods

include/asm-ia64/sn/pci/pcibr_private.h
    some reorg code
    protos for new pcireg_ interfaces

include/asm-ia64/sn/pci/pciio.h
    cleanup

include/asm-ia64/sn/pci/pic.h
    cleanup

include/asm-ia64/sn/sn2/intr.h
    changed IA64_SN2_FIRST_DEVICE_VECTOR and IA64_SN2_LAST_DEVICE_VECTOR

22 years ago[PATCH] Fix x86-64 boot problem
Andrew Morton [Wed, 4 Feb 2004 02:52:16 +0000 (18:52 -0800)]
[PATCH] Fix x86-64 boot problem

From: Andi Kleen <ak@muc.de>

Fix a bug introduced with the last merge that prevented booting with
CONFIG_DEBUG_INFO on on x86-64.  It would corrupt registers in interrupts.
This has hit a few people, so I would consider it as a critical fix.

22 years ago[PATCH] fix memory leak while coredumping
Andrew Morton [Wed, 4 Feb 2004 02:52:07 +0000 (18:52 -0800)]
[PATCH] fix memory leak while coredumping

From: IWAMOTO Toshihiro <iwamoto@valinux.co.jp>

This patch fixes a memory leak that happens when a core file hits the
process's resource limit.

22 years ago[PATCH] SElinux compile fix
Andrew Morton [Wed, 4 Feb 2004 02:51:58 +0000 (18:51 -0800)]
[PATCH] SElinux compile fix

From: Geert Uytterhoeven <geert@linux-m68k.org>

Spinlock code needs <linux/sched.h>

22 years ago[PATCH] DVB compile fix
Andrew Morton [Wed, 4 Feb 2004 02:51:49 +0000 (18:51 -0800)]
[PATCH] DVB compile fix

From: Geert Uytterhoeven <geert@linux-m68k.org>

Never include <asm/delay.h> directly

22 years ago[PATCH] Hisax compile fix
Andrew Morton [Wed, 4 Feb 2004 02:51:40 +0000 (18:51 -0800)]
[PATCH] Hisax compile fix

From: Geert Uytterhoeven <geert@linux-m68k.org>

Never include <asm/delay.h> directly

22 years ago[PATCH] Specialix compile fix
Andrew Morton [Wed, 4 Feb 2004 02:51:31 +0000 (18:51 -0800)]
[PATCH] Specialix compile fix

From: Geert Uytterhoeven <geert@linux-m68k.org>

Compile fix: add missing #include <linux/init.h>

22 years ago[PATCH] Moxa serial compile fixes
Andrew Morton [Wed, 4 Feb 2004 02:51:22 +0000 (18:51 -0800)]
[PATCH] Moxa serial compile fixes

From: Geert Uytterhoeven <geert@linux-m68k.org>

Compile fixes for the Moxa serial drivers:

- Add missing #include <linux/init.h>

- Kill warning if CONFIG_PCI is not set

22 years ago[PATCH] istallion compile fix
Andrew Morton [Wed, 4 Feb 2004 02:51:13 +0000 (18:51 -0800)]
[PATCH] istallion compile fix

From: Geert Uytterhoeven <geert@linux-m68k.org>

Fix compilation if CONFIG_PCI is not set

22 years ago[PATCH] fix compilation warnings in neofb.c
Andrew Morton [Wed, 4 Feb 2004 02:51:04 +0000 (18:51 -0800)]
[PATCH] fix compilation warnings in neofb.c

From: Junio C Hamano <junkio@cox.net>

Compiling neofb.c without MTRR results in two "unused variable" warnings.

22 years ago[PATCH] ppp: try harder to allocate the deflate buffer
Andrew Morton [Wed, 4 Feb 2004 02:50:55 +0000 (18:50 -0800)]
[PATCH] ppp: try harder to allocate the deflate buffer

We're using a 4-order allocation in there and it can fail.  Change it to just
keep reclaiming memory until something gives.

22 years ago[PATCH] Remove the unused kmalloc_percpu_init()
Andrew Morton [Wed, 4 Feb 2004 02:50:46 +0000 (18:50 -0800)]
[PATCH] Remove the unused kmalloc_percpu_init()

From: Martin Hicks <mort@wildopensource.com>

This patch removes kmalloc_percpu_init() from include/linux/percpu.h

It is unused and doesn't seem to be required.

22 years ago[PATCH] missing export of cpu_2_node
Andrew Morton [Wed, 4 Feb 2004 02:50:37 +0000 (18:50 -0800)]
[PATCH] missing export of cpu_2_node

From: Patrick Mansfield <patmans@us.ibm.com>

On NUMAQ, any module which calls __alloc_pages() needs cpu_2_node() (via
cpu_to_node()).

This patch exports cpu_2_node.

22 years ago[PATCH] Remove uneeded dentry assignment
Andrew Morton [Wed, 4 Feb 2004 02:50:28 +0000 (18:50 -0800)]
[PATCH] Remove uneeded dentry assignment

From: James Morris <jmorris@redhat.com>

It seems to me that this dentry assignment in open_namei is not needed per
the patch below.  On this path, dentry is not referenced.

22 years ago[PATCH] CDROMREADAUDIO frames
Andrew Morton [Wed, 4 Feb 2004 02:50:19 +0000 (18:50 -0800)]
[PATCH] CDROMREADAUDIO frames

From: Jens Axboe <axboe@suse.de>

2.6 imposes a 64 frame limit where 2.4 does not (just relies on kmalloc()
failing and limiting frames from that).  That breaks at least on guys app.
With MSF adressing, it's much simpler to be able to ask for a full second
at the time, so I think we should just allow that.  So bump the limit from
64 to CD_FRAMES (which is 75).

22 years ago[PATCH] PCI Scan all functions
Andrew Morton [Wed, 4 Feb 2004 02:50:09 +0000 (18:50 -0800)]
[PATCH] PCI Scan all functions

From: Jake Moilanen <moilanen@austin.ibm.com>

On a ppc64 logically partitioned system, there can be a setup where function
0 of a PCI-PCI bridge is assigned to one partition and (for example) function
2 is assigned to a second partition.  On the second partition, it would
appear that function 0 does not exist, but function 2 does.  If all the
functions are not scanned, everything under function 2 would not be detected.

This patch allows devices that don't respond to function 0, but do respond to
other functions to be marked with a quirk and have all of their functions
scanned.

22 years ago[PATCH] Move cpu_vm_mask to be closer to mmu_context_t in struct mm
Andrew Morton [Wed, 4 Feb 2004 02:49:59 +0000 (18:49 -0800)]
[PATCH] Move cpu_vm_mask to be closer to mmu_context_t in struct mm

From: Jack Steiner <steiner@sgi.com>

The cpu_vm_mask use to be close to the mmu_context_t field in the mm
struct.  Recently some large members were added between "cpu_vm_mask" and
"context".  I suspect that was an oversight.

Here is a patch that puts the fields close together.  This makes it likely
that both fields are in the same cache line.  Since both fields are likely
to be updated at the same time, this may improve performance.

22 years ago[PATCH] Lindent fixed to match reality
Andrew Morton [Wed, 4 Feb 2004 02:49:50 +0000 (18:49 -0800)]
[PATCH] Lindent fixed to match reality

From: Matt Mackall <mpm@selenic.com>

I've been fiddling with cleaning up some old code here and suggest the
following to make Lindent match actual practice more closely. This does:

a) (no -psl)

void *foo(void)
{

 instead of

void *
foo(void) {

b) (no -bs) "sizeof(foo)" rather than "sizeof (foo)"

c) (-ncs) "(void *)foo" rather than "(void *) foo"

22 years ago[PATCH] enable fast symbol lookup via an inverted index in cscope
Andrew Morton [Wed, 4 Feb 2004 02:49:41 +0000 (18:49 -0800)]
[PATCH] enable fast symbol lookup via an inverted index in cscope

From: Louis Zhuang <louis_zhuang@linux.co.intel.com>

enable  fast  symbol lookup via an inverted index.

22 years ago[PATCH] as-iosched.txt update
Andrew Morton [Wed, 4 Feb 2004 02:49:32 +0000 (18:49 -0800)]
[PATCH] as-iosched.txt update

From: Dave Olien <dmo@osdl.org>

acked by npiggin.

22 years ago[PATCH] shrink_list(): check PageSwapCache() after add_to_swap()
Andrew Morton [Wed, 4 Feb 2004 02:49:23 +0000 (18:49 -0800)]
[PATCH] shrink_list(): check PageSwapCache() after add_to_swap()

From: Nikita Danilov <Nikita@Namesys.COM>

shrink_list() checks PageSwapCache() before calling add_to_swap(), this
means that anonymous page that is going to be added to the swap right
now these checks return false and:

 (*) it will be unaccounted for in nr_mapped, and

 (*) it won't be written to the swap if gfp_flags include __GFP_IO but
     not __GFP_FS.

(Both will happen only on the next round of scanning.)

Patch below just moves may_enter_fs initialization down. I am not sure
about (*nr_mapped) increase though.

22 years ago[PATCH] Use address hint in mmap for search
Andrew Morton [Wed, 4 Feb 2004 02:49:14 +0000 (18:49 -0800)]
[PATCH] Use address hint in mmap for search

From: Andi Kleen <ak@suse.de>

When the user gave an address hint in mmap use it as starting point for the
search for !MAP_FIXED.

Currently it is only checked directly and when already used the free area
cache is used as starting point.  With this change you can use mmap(4096,
....) to e.g.  get the lowest free address in your address space, which is
sometimes useful.  For example on x86-64 glibc wants to preferably allocate
thread local data in the first 4GB but use higher addresses when this is
not possible.

This can be a bit more costly in CPU time because it may have to skip over
more VMAs, but gives better semantics for most cases.  Most programs pass
NULL as hint anyways so it won't make any difference for them.

I did it for the generic mmap and for x86-64 for now.  Also minor white
space fixes for x86-64.

22 years ago[PATCH] rate limit nr_free_pages
Andrew Morton [Wed, 4 Feb 2004 02:49:05 +0000 (18:49 -0800)]
[PATCH] rate limit nr_free_pages

From: Jes Sorensen <jes@trained-monkey.org>

nr_free_pages() is expensive, especially on large SMP machines.  The patch
changes the memory overcommit code so that it only calls nr_free_pages() is
we're about to fail the allocation attempt.

22 years ago[PATCH] ide-cd mo write protect
Andrew Morton [Wed, 4 Feb 2004 02:48:56 +0000 (18:48 -0800)]
[PATCH] ide-cd mo write protect

From: Jens Axboe <axboe@suse.de>

It's from Pascal Schmidt and adds write protect handling to ide-cd along
with support for non-2kb block sizes.

22 years ago[PATCH] proc_check_root() locking fix
Andrew Morton [Wed, 4 Feb 2004 02:48:48 +0000 (18:48 -0800)]
[PATCH] proc_check_root() locking fix

From: Maneesh Soni <maneesh@in.ibm.com>

The patch fixes locking in proc_check_root().  It brings is_subdir() call
under vfsmount_lock.  Holding vfsmount_lock will ensure mnt_mountpoint
dentry is intact and the dentry does not go away while it is being checked
in is_subdir().

22 years ago[PATCH] is_subdir locking fix
Andrew Morton [Wed, 4 Feb 2004 02:48:39 +0000 (18:48 -0800)]
[PATCH] is_subdir locking fix

From: Maneesh Soni <maneesh@in.ibm.com>

o The following patch fixes is_subdir() races with d_move. Due to concurrent
  d_move, in is_subdir() we can end up accessing freed d_parent pointer in
  case of pre-emptible kernel. To avoid this we can use rcu_read_lock() and
  rcu_read_unlock().

o This also fixes the seqlock uses in is_subdir() as we need to restart the
  the inner loop with the origianl new_dentry passed to the routine in case
  of any rename occured while we are traversing d_parent links.

22 years ago[PATCH] ac97 OSS driver removal fix
Andrew Morton [Wed, 4 Feb 2004 02:48:30 +0000 (18:48 -0800)]
[PATCH] ac97 OSS driver removal fix

From: Andrew Zabolotny <zap@homelink.ru>

ac97_unregister_driver() is nulling out the ->driver field for all codecs.
It should only null the codecs which are using this driver.

22 years ago[PATCH] u_int32_t causes cross-compile problems
Andrew Morton [Wed, 4 Feb 2004 02:48:21 +0000 (18:48 -0800)]
[PATCH] u_int32_t causes cross-compile problems

From: Pratik Solanki <pratik.solanki@timesys.com>

I came across this C standards issue while cross-compiling the Linux kernel
with gcc on Solaris.  The file gen_crc32table.c uses the non-standard type
u_int32_t.  It's possible that the host machine's sys/types.h does not
define u_int32_t.  The attached patch replaces u_int32_t with the POSIX
standard uint32_t and includes POSIX inttypes.h instead of sys/types.h.