]> git.hungrycats.org Git - linux/log
linux
22 years ago[PATCH] ia64: sn2 defconfig file
Jesse Barnes [Fri, 19 Dec 2003 07:40:20 +0000 (23:40 -0800)]
[PATCH] ia64: sn2 defconfig file

As promised, here's a patch to add an sn2 defconfig file to get people
started with 2.6 kernels.  I even turned on CONFIG_IA64_SGI_SIM support
to make Jack happy :)

22 years ago[PATCH] ia64: initialize bootmem maps in reverse order
Jesse Barnes [Fri, 19 Dec 2003 07:35:51 +0000 (23:35 -0800)]
[PATCH] ia64: initialize bootmem maps in reverse order

The arch-independent bootmem code now requires that arches initialize
their bootmem maps in reverse order (in particular, from high to low
addesses), otherwise alloc_bootmem_pages_low() won't work.  This change
makes the ia64 code do just that, so that machines without an IOMMU can
allocate their bounce buffers in low memory at early boot.  It also adds
a sanity check to the early init code to make sure that each node has a
local data area, because if they don't, many things will break later on
and may be hard to track down.

22 years ago[PATCH] ia64: Prevent SAL calls from being preempted
Bjorn Helgaas [Fri, 19 Dec 2003 07:20:29 +0000 (23:20 -0800)]
[PATCH] ia64: Prevent SAL calls from being preempted

(SAL_CALL_REENTRANT): Disable preemption around the SAL call to
make sure we don't get rescheduled on a different CPU.

22 years ago[PATCH] ia64: Force generic and hp kernels to use 16MB granules
Bjorn Helgaas [Fri, 19 Dec 2003 07:19:55 +0000 (23:19 -0800)]
[PATCH] ia64: Force generic and hp kernels to use 16MB granules

This forces the granule size to 16MB for HP zx1 and generic
kernels.  HP sx1000 machines require this.

22 years ago[PATCH] ia64: Remove unused ACPI functions.
Bjorn Helgaas [Fri, 19 Dec 2003 07:18:47 +0000 (23:18 -0800)]
[PATCH] ia64: Remove unused ACPI functions.

Remove unused functions:
    acpi_get_prt()
    acpi_get_interrupt_model()
    acpi_get_addr_space()

22 years ago[PATCH] ia64: Remove extraneous printks (we get the same information from ACPI).
Bjorn Helgaas [Fri, 19 Dec 2003 07:17:48 +0000 (23:17 -0800)]
[PATCH] ia64: Remove extraneous printks (we get the same information from ACPI).

(iosapic_init): Remove extraneous printk.
(pci_acpi_scan_root): Remove extraneous printk.

22 years ago[PATCH] ia64: Fix PCI root bridge resources to handle prior allocations.
Bjorn Helgaas [Fri, 19 Dec 2003 07:16:43 +0000 (23:16 -0800)]
[PATCH] ia64: Fix PCI root bridge resources to handle prior allocations.

(alloc_resources):  Use insert_resource(), not request_resource(), to
allocate PCI root bridge windows.  This fixes the problem
where root bridge window allocation fails because an early
driver (like VGA) has already allocated things.

22 years ago[PATCH] ia64: make NODES_SHIFT a little biggger
Jesse Barnes [Fri, 19 Dec 2003 07:15:20 +0000 (23:15 -0800)]
[PATCH] ia64: make NODES_SHIFT a little biggger

Make NODES_SHIFT larger to accomodate 256 node machines.

22 years ago[PATCH] ia64: update sn2 MAINTAINERS file entry
Jesse Barnes [Fri, 19 Dec 2003 07:14:29 +0000 (23:14 -0800)]
[PATCH] ia64: update sn2 MAINTAINERS file entry

22 years ago[PATCH] ia64: make cpu_to_node_map unsigned
Jesse Barnes [Fri, 19 Dec 2003 07:13:48 +0000 (23:13 -0800)]
[PATCH] ia64: make cpu_to_node_map unsigned

This small fix is needed for machines with more than 128 nodes.

22 years ago[PATCH] ia64: prevent buffer-overrun in acpi_numa_memory_affinity_init()
Jack Steiner [Fri, 19 Dec 2003 07:09:49 +0000 (23:09 -0800)]
[PATCH] ia64: prevent buffer-overrun in acpi_numa_memory_affinity_init()

The code in acpi_numa_memory_affinity_init that sorts the node_memblk
can overrun the array & clobber the memory that follows the end of the
array. The error will be seen only on systems that fill the
node_memblk array and only if SAL doesnt sort the entries in the SRAT.

22 years ago[PATCH] ia64: enable out-of-tree compilation for IA64
Peter Chubb [Fri, 19 Dec 2003 03:48:09 +0000 (19:48 -0800)]
[PATCH] ia64: enable out-of-tree compilation for IA64

22 years agoMerge tiger.hpl.hp.com:/data1/bk/vanilla/linux-2.5
David Mosberger [Wed, 17 Dec 2003 14:48:35 +0000 (06:48 -0800)]
Merge tiger.hpl.hp.com:/data1/bk/vanilla/linux-2.5
into tiger.hpl.hp.com:/data1/bk/lia64/to-linus-2.5

22 years agoLinux 2.6.0 v2.6.0
Linus Torvalds [Wed, 17 Dec 2003 10:56:13 +0000 (02:56 -0800)]
Linux 2.6.0

22 years ago[PATCH] qla1280 crash fix in error handling
Jes Sorensen [Wed, 17 Dec 2003 00:26:43 +0000 (16:26 -0800)]
[PATCH] qla1280 crash fix in error handling

This fixes a bug in the qla1280 driver where it would leave a pointer to
an on the stack completion event in a command structure if
qla1280_mailbox_command fails.  The result is that the interrupt handler
later tries to complete() garbage on the stack.  The mailbox command can
fail if a device on the bus decides to lock up etc.

22 years ago[PATCH] CDROM_SEND_PACKET bug
Jens Axboe [Wed, 17 Dec 2003 00:12:29 +0000 (16:12 -0800)]
[PATCH] CDROM_SEND_PACKET bug

I just found Yet Another Bug in scsi_ioctl - CDROM_SEND_PACKET puts a
kernel pointer in hdr->cmdp, where sg_io() expects to find user address.
This worked up until recently because of the memcpy bug, but now it
doesn't because we do the proper copy_from_user().

This fix undoes the user copy code from sg_io, and instead makes the
SG_IO ioctl copy it locally.  This makes SG_IO and CDROM_SEND_PACKET
agree on the calling convention, and everybody is happy.

I've tested that both

   cdrecord -dev=/dev/hdc -inq

and

   cdrecord -dev=ATAPI:/dev/hdc -inq

works now.  The former will use SG_IO, the latter CDROM_SEND_PACKET (and
incidentally would work in both 2.4 and 2.6, if it wasn't for
CDROM_SEND_PACKET sucking badly in 2.4).

22 years agoia64: Bring export of spin-lock contention-routines in sync with
David Mosberger [Tue, 16 Dec 2003 01:17:33 +0000 (17:17 -0800)]
ia64: Bring export of spin-lock contention-routines in sync with
this change:  Jim Wilson says that gcc v3.3 also supports
marking ar.pfs as clobbered, so use ia64_spinlock_contention()
for any GCC with v3.3 or newer.

22 years ago[PATCH] Fix IDE bus reset and DMA disable when reading blank DVD-R
Jens Axboe [Mon, 15 Dec 2003 23:51:55 +0000 (15:51 -0800)]
[PATCH] Fix IDE bus reset and DMA disable when reading blank DVD-R

From Jon Burgess:

  There is a problems with blank DVD media using the ide-cd driver.

  When we attempt to read the blank disk, the drive responds to the read
  request by returning a "blank media" error.  The kernel doesn't have
  any special case handling for this sense value and retries the request
  a couple of times, then gives up and does a bus reset and disables DMA
  to the device.

  Which obviously doesn't help the situation.

  The sense key value of 8 isn't listed in ide-cd.h, but it is listed in
  scsi.h as a "BLANK_CHECK" error.

  This trivial patch treats this error condition as a reason to abort
  the request.  This behaviour is the same as what we do with a blank CD-R.

  It looks like the same fix might be desired for 2.4 as well, although
  is perhaps not so important since scsi-ide is normally used instead.

22 years agoMerge bk://linuxusb.bkbits.net/gregkh-2.6
Linus Torvalds [Mon, 15 Dec 2003 13:25:24 +0000 (05:25 -0800)]
Merge bk://linuxusb.bkbits.net/gregkh-2.6
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[PATCH] Fix possible bio corruption with RAID5
Neil Brown [Mon, 15 Dec 2003 08:35:56 +0000 (00:35 -0800)]
[PATCH] Fix possible bio corruption with RAID5

 1/ make sure raid5 doesn't try to handle multiple overlaping
    requests at the same time as this would confuse things badly.
    Currently it justs BUGs if this is attempted.
 2/ Fix a possible data-loss-on-write problem.  If two or
    more bio's that write to the same page are processed at the
    same time, only the first was actually commited to storage.
 3/ Fix a use-after-free bug.  raid5 keeps the bio's it is given
    in linked lists when more than one bio touch a single page.
    In some cases the tail of this list can be freed, and
    the current test for 'are we at the end' isn't reliable.
    This patch strengths the test to make it reliable.

22 years agoFix thread group leader zombie leak
Linus Torvalds [Sun, 14 Dec 2003 10:28:27 +0000 (02:28 -0800)]
Fix thread group leader zombie leak

Petr Vandrovec noticed a problem where the thread group leader
would not be properly reaped if the parent of the thread group
was ignoring SIGCHLD, and the thread group leader had exited
before the last sub-thread.

Fixed by Ingo Molnar.

22 years agoMore subtle SMP bugs in prepare_to_wait()/finish_wait().
Linus Torvalds [Sat, 13 Dec 2003 13:36:30 +0000 (05:36 -0800)]
More subtle SMP bugs in prepare_to_wait()/finish_wait().

This time we have a SMP memory ordering issue in prepare_to_wait(),
where we really need to make sure that subsequent tests for the
event we are waiting for can not migrate up to before the wait
queue has been set up.

22 years ago[PATCH] HPFS: missing lock_kernel() in hpfs_readdir()
René Scharfe [Sat, 13 Dec 2003 03:45:10 +0000 (19:45 -0800)]
[PATCH] HPFS: missing lock_kernel() in hpfs_readdir()

In 2.5.x, the BKL was pushed from vfs_readdir() into the filesystem
specific functions.  But only the unlock_kernel() made it into the HPFS
code, lock_kernel() got lost on the way.  This rendered the filesystem
unusable.

This adds the missing lock_kernel().  It's been tested by Timo Maier who
also reported the problem earlier today.

22 years ago[PATCH] no bio unmap on cdb copy failure
Jens Axboe [Sat, 13 Dec 2003 03:42:12 +0000 (19:42 -0800)]
[PATCH] no bio unmap on cdb copy failure

The previous scsi_ioctl.c patch didn't cleanup the buffer/bio in the
error case.

Fix it by copying the command data earlier.

22 years agoia64: Fix compiler warning in intrinsics.h.
David Mosberger [Fri, 12 Dec 2003 17:04:41 +0000 (09:04 -0800)]
ia64: Fix compiler warning in intrinsics.h.

22 years agoia64: Fix ivt overflow that occurred when turning on
David Mosberger [Fri, 12 Dec 2003 16:56:49 +0000 (08:56 -0800)]
ia64: Fix ivt overflow that occurred when turning on
CONFIG_DISABLE_VHPT.

22 years agoia64: Fix bug discovered by Bill Nottingham & Jakub Jelinek where
David Mosberger [Fri, 12 Dec 2003 16:04:17 +0000 (08:04 -0800)]
ia64: Fix bug discovered by Bill Nottingham & Jakub Jelinek where
put_user() arguments with function-calls would cause the
macro to return unexpected values.  Fixed by avoiding macro
argument evaluation while r8/r9 are in use for exception-handling
purposes.  Also, consolidated access-macros so that GCC and
Intel compiler use 90% the same code.

22 years agoFix subtle bug in "finish_wait()", which can cause kernel stack
Linus Torvalds [Fri, 12 Dec 2003 06:20:08 +0000 (22:20 -0800)]
Fix subtle bug in "finish_wait()", which can cause kernel stack
corruption on SMP because of another CPU still accessing a waitqueue
even after it was de-allocated.

Use a careful version of the list emptiness check to make sure we
don't de-allocate the stack frame before the waitqueue is all done.

22 years agoia64: Based on patch by Jerome Marchand: Add ia64-optimized
David Mosberger [Thu, 11 Dec 2003 04:22:16 +0000 (20:22 -0800)]
ia64: Based on patch by Jerome Marchand: Add ia64-optimized
atomic_dec_and_lock().  Actually, this could be the generic
version for any platform that has cmpxchg().

22 years agokobject: fix bug where a parent could be deleted before a child device.
Greg Kroah-Hartman [Thu, 11 Dec 2003 01:49:24 +0000 (17:49 -0800)]
kobject: fix bug where a parent could be deleted before a child device.

22 years agoia64: Switch places for the gate pages and the guard page. This improves
David Mosberger [Wed, 10 Dec 2003 09:34:27 +0000 (01:34 -0800)]
ia64: Switch places for the gate pages and the guard page.  This improves
backwards-compatibility with older (broken) versions of GCC which
recognize a signal-handler only if it is in the address range
from 0xa000000000000100. to 0xa000000000020000.

22 years agoia64: Jim Wilson says that gcc v3.3 also supports marking ar.pfs as
David Mosberger [Wed, 10 Dec 2003 09:29:05 +0000 (01:29 -0800)]
ia64: Jim Wilson says that gcc v3.3 also supports marking ar.pfs as
clobbered, so use ia64_spinlock_contention() for any GCC with
v3.3 or newer.

22 years agoMerge kroah.com:/home/greg/linux/BK/bleed-2.5
Greg Kroah-Hartman [Wed, 10 Dec 2003 08:43:42 +0000 (00:43 -0800)]
Merge kroah.com:/home/greg/linux/BK/bleed-2.5
into kroah.com:/home/greg/linux/BK/usb-2.6

22 years ago[PATCH] USB: prevent catch-all USB aliases in modules.alias
Andrey Borzenkov [Wed, 10 Dec 2003 08:00:53 +0000 (00:00 -0800)]
[PATCH] USB: prevent catch-all USB aliases in modules.alias

visor.c defines one empty slot in USB ids table that can be filled in at
runtime using module parameters. file2alias generates catch-all alias for it:

alias usb:v*p*dl*dh*dc*dsc*dp*ic*isc*ip* visor

patch adds the same sanity check as in depmod to scripts/file2alias.

22 years ago[PATCH] USB: fix bug for multiple opens on ttyUSB devices.
Greg Kroah-Hartman [Wed, 10 Dec 2003 06:39:48 +0000 (22:39 -0800)]
[PATCH] USB: fix bug for multiple opens on ttyUSB devices.

This patch fixes the bug where running ppp over a ttyUSB device would fail.

22 years ago[PATCH] USB: Fix connect/disconnect race
Herbert Xu [Tue, 9 Dec 2003 09:40:44 +0000 (01:40 -0800)]
[PATCH] USB: Fix connect/disconnect race

This patch was integrated by you in 2.4 six months ago.  Unfortunately
it never got into 2.5.  Without it you can end up with crashes such
as http://bugs.debian.org/218670

22 years ago[PATCH] USB: fix bug not setting device state following usb_device_reset()
Alan Stern [Tue, 9 Dec 2003 07:41:57 +0000 (23:41 -0800)]
[PATCH] USB: fix bug not setting device state following usb_device_reset()

22 years ago[PATCH] USB: fix race with signal delivery in usbfs
Oliver Neukum [Tue, 9 Dec 2003 02:18:07 +0000 (18:18 -0800)]
[PATCH] USB: fix race with signal delivery in usbfs

apart from locking bugs, there are other races. This fixes one with
signal delivery. The signal should be delivered _before_ the reciever
is woken.

22 years ago[PATCH] USB: fix sleping in interrupt bug in auerswald driver
Oliver Neukum [Tue, 9 Dec 2003 02:00:49 +0000 (18:00 -0800)]
[PATCH] USB: fix sleping in interrupt bug in auerswald driver

this fixes two instances of GFP_KERNEL from completion handlers.

22 years ago[PATCH] USB: mark the scanner driver as obsolete
Tom Rini [Tue, 9 Dec 2003 01:42:34 +0000 (17:42 -0800)]
[PATCH] USB: mark the scanner driver as obsolete

On Mon, Dec 01, 2003 at 11:21:58AM -0800, Greg KH wrote:
> Can't you use xsane without the scanner kernel driver?  I thought the
> latest versions used libusb/usbfs to talk directly to the hardware.
> Because of this, the USB scanner driver is marked to be removed from the
> kernel sometime in the near future.

After a bit of mucking around (and possibly finding a bug with debian's
libusb/xsane/hotplug interaction, nothing seems to run
/etc/hotplug/usb/libusbscanner and thus only root can scan, anyone whose
got this working please let me know), the problem does not exist if I
only use  libusb xsane.

How about the following:

22 years ago[PATCH] USB storage: fix for jumpshot and datafab devices
Matthew Dharm [Tue, 9 Dec 2003 01:35:37 +0000 (17:35 -0800)]
[PATCH] USB storage: fix for jumpshot and datafab devices

This patch fixes some obvious errors in the jumpshot and datafab drivers.

This should close out Bugzilla bug #1408

> Date: Mon, 1 Dec 2003 12:14:53 -0500 (EST)
> From: Alan Stern <stern@rowland.harvard.edu>
> Subject: Patch from Eduard Hasenleithner
> To: Matthew Dharm <mdharm-usb@one-eyed-alien.net>
> cc: USB Storage List <usb-storage@one-eyed-alien.net>
>
> Matt:
>
> Did you see this patch?  It was posted to the usb-development mailing list
> about a week ago, before I started making all my changes.  It is clearly
> correct and necessary.
>
> Alan Stern

22 years ago[PATCH] scsi_ioctl memcpy'ing user address
Jens Axboe [Tue, 9 Dec 2003 01:03:05 +0000 (17:03 -0800)]
[PATCH] scsi_ioctl memcpy'ing user address

James reported a bug in scsi_ioctl.c where it mem copies a user pointer
instead of using copy_from_user(). I inadvertently introduced this one
when getting rid of CDROM_SEND_PACKET. Here's a trivial patch to fix it.

22 years ago[IPV6]: Fix TCP socket leak.
Arnaldo Carvalho de Melo [Mon, 8 Dec 2003 13:40:01 +0000 (05:40 -0800)]
[IPV6]: Fix TCP socket leak.

TCP IPV6 ->hash() method should not grab a socket reference.

22 years ago[PATCH] ia64: make perfmon CONFIG_PREEMPT-safe again
Peter Chubb [Mon, 8 Dec 2003 09:37:01 +0000 (01:37 -0800)]
[PATCH] ia64: make perfmon CONFIG_PREEMPT-safe again

Here's a  fix for non-preemption safety in perfmon.c.

I haven't tried it while running a preemption stress test, but this
allows q-syscollect to work.

22 years ago[PATCH] USB: fix race with hub devices disconnecting while stuff is still happening...
Greg Kroah-Hartman [Mon, 8 Dec 2003 05:58:14 +0000 (21:58 -0800)]
[PATCH] USB: fix race with hub devices disconnecting while stuff is still happening to them.

22 years ago[PATCH] USB: fix remove device after set_configuration
David Brownell [Mon, 8 Dec 2003 05:28:46 +0000 (21:28 -0800)]
[PATCH] USB: fix remove device after set_configuration

If a device can't be configured, the current test9 code forgets
to clean it out of sysfs.  This resolves that issue, so the retry
in usb_new_device() stands a chance of working.

The enumeration code still doesn't handle such errors well, but
at least this way that hub port can be used for another device.

22 years ago[PATCH] USB: register usb-serial ports in the proper place in sysfs
Greg Kroah-Hartman [Mon, 8 Dec 2003 03:10:51 +0000 (19:10 -0800)]
[PATCH] USB: register usb-serial ports in the proper place in sysfs

They should be bound to the interface the driver is attached to, not
the device.

22 years ago[PATCH] tmpfs oops fix
James McMechan [Sun, 7 Dec 2003 13:57:40 +0000 (05:57 -0800)]
[PATCH] tmpfs oops fix

The problem was that the cursor was in the list being walked, and when
the pointer pointed to the cursor the list_del/list_add_tail pair would
oops trying to find the entry pointed to by the prev pointer of the
deleted cursor element.

The solution I found was to move the list_del earlier, before the
beginning of the list walk. since it is not used during the list walk and
should not count in the list enumeration it can be deleted, then the
list pointer cannot point to it so it can be added safely with the
list_add_tail without oopsing, and everything works as expected.

I am unable to oops this version with any of my test programs.

Patch acked by Al Viro.

22 years agoMerge master.kernel.org:/home/davem/BK/net-2.5
Linus Torvalds [Sun, 7 Dec 2003 12:20:28 +0000 (04:20 -0800)]
Merge master.kernel.org:/home/davem/BK/net-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[PKT_SCHED]: In HTB, filters must be destroyed before the classes.
Martin Devera [Sun, 7 Dec 2003 11:33:36 +0000 (03:33 -0800)]
[PKT_SCHED]: In HTB, filters must be destroyed before the classes.

22 years ago[PKT_SCHED]: Do not dereference the special pointer value 'HTB_DIRECT'.
David S. Miller [Sun, 7 Dec 2003 11:31:37 +0000 (03:31 -0800)]
[PKT_SCHED]: Do not dereference the special pointer value 'HTB_DIRECT'.

Based upon a patch from devik.

22 years ago[PATCH] wireless airo oops fix
Jeff Garzik [Sun, 7 Dec 2003 03:00:48 +0000 (19:00 -0800)]
[PATCH] wireless airo oops fix

From Javier Achirica:

Delay MIC activation to prevent Oops

22 years agoFix the PROT_EXEC breakage on anonymous mmap.
Linus Torvalds [Sat, 6 Dec 2003 06:34:40 +0000 (22:34 -0800)]
Fix the PROT_EXEC breakage on anonymous mmap.

Clean up the tests while at it.

22 years ago[PATCH] remove manual driver poisoning of net_device
Jeff Garzik [Fri, 5 Dec 2003 15:35:35 +0000 (07:35 -0800)]
[PATCH] remove manual driver poisoning of net_device

From: Al Viro <viro@parcelfarce.linux.theplanet.co.uk>

   Such poisoning can cause oopses either because the refcount is not
   zero when the poisoning occurs, or due to kernel debugging options
   being enabled.

22 years ago[PATCH] fix oops on unload in pcnet32
Jeff Garzik [Fri, 5 Dec 2003 15:34:00 +0000 (07:34 -0800)]
[PATCH] fix oops on unload in pcnet32

The driver was calling pci_unregister_driver for each _device_, and then
again at the end of the module unload routine.  Remove the call that's
inside the loop, pci_unregister_driver should only be called once.

Caught by Don Fry (and many others)

22 years ago[PATCH] fix use-after-free in libata
Jeff Garzik [Fri, 5 Dec 2003 15:32:14 +0000 (07:32 -0800)]
[PATCH] fix use-after-free in libata

Fixes oops some were seeing on module unload.

Caught by Jon Burgess.

22 years agoMerge master.kernel.org:/home/davem/BK/net-2.5
Linus Torvalds [Fri, 5 Dec 2003 07:28:12 +0000 (23:28 -0800)]
Merge master.kernel.org:/home/davem/BK/net-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[BRIDGE]: Provide correct TOS value to IPv4 routing.
Julian Anastasov [Fri, 5 Dec 2003 04:41:31 +0000 (20:41 -0800)]
[BRIDGE]: Provide correct TOS value to IPv4 routing.

22 years ago[NETFILTER]: In conntrack, do not fragment TSO packets by accident.
David S. Miller [Fri, 5 Dec 2003 04:34:00 +0000 (20:34 -0800)]
[NETFILTER]: In conntrack, do not fragment TSO packets by accident.

22 years ago[PATCH] Fix 'noexec' behaviour
Ulrich Drepper [Thu, 4 Dec 2003 14:26:06 +0000 (06:26 -0800)]
[PATCH] Fix 'noexec' behaviour

We should not allow mmap() with PROT_EXEC on mounts marked "noexec",
since otherwise there is no way for user-supplied executable loaders
(like ld.so and emulator environments) to properly honour the
"noexec"ness of the target.

22 years agoMerge bk://linuxusb.bkbits.net/i2c-2.6
Linus Torvalds [Thu, 4 Dec 2003 09:13:08 +0000 (01:13 -0800)]
Merge bk://linuxusb.bkbits.net/i2c-2.6
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[PATCH] ia64: fix typo in vmlinux.lds.S
Jakub Jelínek [Thu, 4 Dec 2003 06:53:04 +0000 (22:53 -0800)]
[PATCH] ia64: fix typo in vmlinux.lds.S

The security init section was incorrectly using PAGE_OFFSET instead of
LOAD_OFFSET.

22 years ago[PATCH] I2C: fix i2c_smbus_write_byte() for i2c-nforce2
Jean Delvare [Thu, 4 Dec 2003 06:14:33 +0000 (22:14 -0800)]
[PATCH] I2C: fix i2c_smbus_write_byte() for i2c-nforce2

This patch fixes i2c_smbus_write_byte() being broken for i2c-nforce2.
This causes trouble when that module is used together with eeprom (which
is also in 2.6). We have had three user reports about the problem.

Credits go to Mark D. Studebaker for finding and fixing the problem.

22 years ago[PATCH] fix broken x86_64 rdtscll
Jens Axboe [Wed, 3 Dec 2003 23:53:31 +0000 (15:53 -0800)]
[PATCH] fix broken x86_64 rdtscll

The scheduler is completed b0rked on x86_64, and I finally found out
why.  sched_clock() always returned 0, because rdtscll() always returned
0.  The 'a' in the macro doesn't agree with the 'a' in the function,
yippe :-)

This is a show stopper for x86_64.

22 years agoMerge master.kernel.org:/home/davem/BK/net-2.5
Linus Torvalds [Wed, 3 Dec 2003 05:36:08 +0000 (21:36 -0800)]
Merge master.kernel.org:/home/davem/BK/net-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[PATCH] Fix /proc access to dead thread group list oops
Ingo Molnar [Wed, 3 Dec 2003 04:59:12 +0000 (20:59 -0800)]
[PATCH] Fix /proc access to dead thread group list oops

The pid_alive() check within the loop is incorrect.  If we are within
the tasklist lock and the thread group leader is valid then the thread
chain will be fully intact.

Instead, the check should be _outside_ the loop, since if the group
leader no longer exists, the whole list is gone and we must not try
to access it.

Move the check around, and add comment.

Bug-hunting and fix by Srivatsa Vaddagiri

22 years ago[RTNETLINK]: Add RTPROT_XORP.
Pavlin Radoslavov [Wed, 3 Dec 2003 03:57:38 +0000 (19:57 -0800)]
[RTNETLINK]: Add RTPROT_XORP.

22 years ago[NETFILTER]: Sanitize ip_ct_tcp_timeout_close_wait value, from 2.4.x
Harald Welte [Wed, 3 Dec 2003 03:57:05 +0000 (19:57 -0800)]
[NETFILTER]: Sanitize ip_ct_tcp_timeout_close_wait value, from 2.4.x

22 years ago[IPV6]: Fix ipv4 mapped address calculation in udpv6_sendmsg().
Hideaki Yoshifuji [Wed, 3 Dec 2003 03:56:26 +0000 (19:56 -0800)]
[IPV6]: Fix ipv4 mapped address calculation in udpv6_sendmsg().

22 years agoFix ide-scsi.c uninitialized variable
Linus Torvalds [Tue, 2 Dec 2003 11:04:01 +0000 (03:04 -0800)]
Fix ide-scsi.c uninitialized variable

22 years agoFix x86 kernel page fault error codes
Linus Torvalds [Tue, 2 Dec 2003 10:41:48 +0000 (02:41 -0800)]
Fix x86 kernel page fault error codes

22 years ago[PATCH] Fix lost wakeups problem
Ingo Molnar [Mon, 1 Dec 2003 08:51:30 +0000 (00:51 -0800)]
[PATCH] Fix lost wakeups problem

When doing sync wakeups we must not skip the notification of other cpus
if the task is not on this runqueue.

22 years ago[PATCH] ia64: fix samp_affinity user-space accesses
Paul Jackson [Mon, 1 Dec 2003 05:13:14 +0000 (21:13 -0800)]
[PATCH] ia64: fix samp_affinity user-space accesses

Here is a new improved patch for verifying user access to string
passed in to kernel on write to /proc/irq/<pid>/smp_affinity.

The access_ok() but missing __get_user() on each byte earlier patch
has been replaced with a copy_from_user().

I have built it and verified that it handles write requests
as before, on an ia64 system (well - you can no longer pass
more than 14 spaces after the 'R' - tough).

22 years ago[PATCH] Missing initialization of /proc/net/tcp seq_file
Hirofumi Ogawa [Mon, 1 Dec 2003 02:40:47 +0000 (18:40 -0800)]
[PATCH] Missing initialization of /proc/net/tcp seq_file

We need to initialize st->state in tcp_seq_start().  Otherwise
tcp_seq_stop() is run with previous st->state, and it calls the unneeded
unlock etc, causing a kernel crash.

22 years agoia64: Fix a bug in sigtramp() which corrupted ar.rnat when unwinding
David Mosberger [Wed, 26 Nov 2003 07:03:24 +0000 (23:03 -0800)]
ia64: Fix a bug in sigtramp() which corrupted ar.rnat when unwinding
across a signal trampoline (in user space).  Reported by
Laurent Morichetti.

22 years agoLinux 2.6.0-test11 v2.6.0-test11
Linus Torvalds [Wed, 26 Nov 2003 04:38:22 +0000 (20:38 -0800)]
Linux 2.6.0-test11

22 years ago[PATCH] Lastminute IEEE-1394 fixes
Ben Collins [Wed, 26 Nov 2003 04:15:46 +0000 (20:15 -0800)]
[PATCH] Lastminute IEEE-1394 fixes

I've got a lot more changes than what's included here.  I've put this
down to the bear minimum to get things working sanely.

Mainly, I just want to get all the people hit by this a chance to use
2.6.0 without having to get our tree. Changes itemized:

 - Fix deadlock possibility in csr.c:read_maps()
 - Fix kmalloc to use ATOMIC in highlevel.c.
 - s/in_interrupt/irqs_disabled/ in ieee1394_transactions.c to fix
   warnings when transactions occured.
 - Introduce a release callback for the host driver and use it correctly.
 - Reorganize the nodemgr probe so we do an initial scan to discover
   devices, check IRM/CycleMaster, then do a final full probe when things
   are kosher. Fixes a problem where device registration and hotplug
   would cause some serious problems when a bus reset was forced in the
   middle of the probe.

22 years agoMerge master.kernel.org:/home/davem/BK/sparc-2.5
Linus Torvalds [Tue, 25 Nov 2003 23:06:55 +0000 (15:06 -0800)]
Merge master.kernel.org:/home/davem/BK/sparc-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years agoMerge master.kernel.org:/home/davem/BK/net-2.5
Linus Torvalds [Tue, 25 Nov 2003 23:04:46 +0000 (15:04 -0800)]
Merge master.kernel.org:/home/davem/BK/net-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years agoMerge http://lia64.bkbits.net/to-linus-2.5
Linus Torvalds [Tue, 25 Nov 2003 23:02:29 +0000 (15:02 -0800)]
Merge http://lia64.bkbits.net/to-linus-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[PATCH] prevent oops from read of proc entry for tty drivers
Stephen Hemminger [Tue, 25 Nov 2003 23:01:31 +0000 (15:01 -0800)]
[PATCH] prevent oops from read of proc entry for tty drivers

There are /proc handles there setup by proc_tty_register_driver, but there is
no module ownership association, so anything that reads after module unload
will blow.

The trivial fix is to propagate the owner of tty_driver to proc entry.

22 years ago[PATCH] fix smsc-ircc2.c double free
Jan Marek [Tue, 25 Nov 2003 23:00:04 +0000 (15:00 -0800)]
[PATCH] fix smsc-ircc2.c double free

This fixes a small, but important bug in smsc-ircc2.c.

Without this the driver tries release the same region twice.

22 years agoia64: Make core-dumps work even when executing in the gate DSO.
David Mosberger [Tue, 25 Nov 2003 15:26:14 +0000 (07:26 -0800)]
ia64: Make core-dumps work even when executing in the gate DSO.
      Tracked down by Rohit Seth.

22 years agoia64: Correct FIXADDR_USER_END so that single-stepping in the gate DSO works again.
David Mosberger [Tue, 25 Nov 2003 15:08:23 +0000 (07:08 -0800)]
ia64: Correct FIXADDR_USER_END so that single-stepping in the gate DSO works again.

22 years ago[IRDA]: Fix SKB leaks.
Jean Tourrilhes [Tue, 25 Nov 2003 10:53:30 +0000 (02:53 -0800)]
[IRDA]: Fix SKB leaks.

Do not special case ENOMEM as an error code, we should free up
the packet for all errors.

22 years ago[SPARC]: Fix build failures in IGA frame buffer introduced by janitor changes.
Pete Zaitcev [Tue, 25 Nov 2003 09:52:11 +0000 (01:52 -0800)]
[SPARC]: Fix build failures in IGA frame buffer introduced by janitor changes.

Most of the breakage came from cset 1.838.10.3

22 years ago[SPARC]: Add stub sched_clock() implementation.
Pete Zaitcev [Tue, 25 Nov 2003 09:49:02 +0000 (01:49 -0800)]
[SPARC]: Add stub sched_clock() implementation.

22 years ago[BRIDGE]: Fix netfilter config tests.
Bart De Schuymer [Tue, 25 Nov 2003 09:16:13 +0000 (01:16 -0800)]
[BRIDGE]: Fix netfilter config tests.

The br_netfilter_{init,fini}() calls need to be guarded by
BRIDGE_NETFILTER, from Michal Janusz Miroslaw.

22 years ago[IPV6]: Redo stateless addrconf properly.
Hideaki Yoshifuji [Tue, 25 Nov 2003 09:10:18 +0000 (01:10 -0800)]
[IPV6]: Redo stateless addrconf properly.

When bringing an interface down do not forget to clear the
IF_RS_SENT and IF_RA_RCVD flags.

22 years agoMerge bk://gkernel.bkbits.net/libata-2.5
Linus Torvalds [Tue, 25 Nov 2003 07:18:43 +0000 (23:18 -0800)]
Merge bk://gkernel.bkbits.net/libata-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[PATCH] JFS: Avoid segfault when dirty inodes are written on readonly mount
Dave Kleikamp [Tue, 25 Nov 2003 05:41:19 +0000 (21:41 -0800)]
[PATCH] JFS: Avoid segfault when dirty inodes are written on readonly mount

This fixes an oops that can occur if JFS is used as the root filesystem.

Writes to a device node may cause a ->write_inode to be called during a
read-only mount.  JFS needs to check for NULL log in jfs_flush_journal.

22 years agoFix error return on concurrent fork() with threaded exit()
Linus Torvalds [Tue, 25 Nov 2003 02:46:32 +0000 (18:46 -0800)]
Fix error return on concurrent fork() with threaded exit()

22 years agoMerge master.kernel.org:/home/davem/BK/net-2.5
Linus Torvalds [Mon, 24 Nov 2003 12:48:39 +0000 (04:48 -0800)]
Merge master.kernel.org:/home/davem/BK/net-2.5
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[ECONET]: Do not leak SKBs if ec_queue_packet() fails.
David S. Miller [Mon, 24 Nov 2003 12:00:19 +0000 (04:00 -0800)]
[ECONET]: Do not leak SKBs if ec_queue_packet() fails.

Also, make sure NET_RX_DROP is returned if we did not accept the
packet.

22 years ago[PPPOE]: Do not leak SKB if sock_queue_rcv_skb() fails.
David S. Miller [Mon, 24 Nov 2003 11:49:48 +0000 (03:49 -0800)]
[PPPOE]: Do not leak SKB if sock_queue_rcv_skb() fails.

22 years ago[NET]: In sock_queue_rcv_skb(), do not deref skb->len after it is queued to the socket.
David S. Miller [Mon, 24 Nov 2003 11:44:51 +0000 (03:44 -0800)]
[NET]: In sock_queue_rcv_skb(), do not deref skb->len after it is queued to the socket.

In implementations that use no socket locking, such as RAW sockets,
once we queue the SKB to the socket another cpu can remove the SKB
from the socket queue and free up the SKB making the skb->len access
touch freed memory.

Based upon a report from Burton Windle, kernel bugzilla #937

22 years agoMerge bk://linux-scsi.bkbits.net/scsi-bugfixes-2.6
Linus Torvalds [Mon, 24 Nov 2003 10:56:37 +0000 (02:56 -0800)]
Merge bk://linux-scsi.bkbits.net/scsi-bugfixes-2.6
into home.osdl.org:/home/torvalds/v2.5/linux

22 years ago[libata] bump versions for core and serverworks driver
Jeff Garzik [Mon, 24 Nov 2003 06:26:26 +0000 (01:26 -0500)]
[libata] bump versions for core and serverworks driver

22 years ago[libata] Fix flush of Device Control register to device
Benjamin Herrenschmidt [Mon, 24 Nov 2003 06:19:35 +0000 (01:19 -0500)]
[libata] Fix flush of Device Control register to device

Also add some ServerWorks-specific tweaks.

22 years agoMerge redhat.com:/spare/repo/linux-2.5
Jeff Garzik [Mon, 24 Nov 2003 05:33:55 +0000 (00:33 -0500)]
Merge redhat.com:/spare/repo/linux-2.5
into redhat.com:/spare/repo/libata-2.5